career-ops-hq/career-ops
52.5
Adequate · 29 October 2026
208.4k
lines of production code
JavaScript
with TypeScript
4
measurements over time
What this system is
Career-ops is a local-first, AI-driven job search and application management system that automates the discovery, evaluation, and submission of job applications. It features a plugin-based architecture for ingesting leads from diverse sources, an agentic assistant for driving browser-based application forms, and a comprehensive dashboard for tracking pipeline analytics and managing follow-ups. The system operates primarily through a CLI and an optional web UI, keeping all data and processing local to ensure privacy and security.
Features
API endpoint to override or clear the next follow-up date
A new API route at \/api/followups/override\ allows users to programmatically set or remove a specific next follow-up date for an application. The POST endpoint accepts an application number and a valid ISO date to replace the current pin, while the DELETE endpoint removes the pin for a given application. Both operations are protected by a file-based lock to prevent race conditions with the background seeder process, and the endpoint returns 400 errors for malformed JSON or invalid inputs rather than 500 server errors.
web/src/app/api/followups/override · high confidence
Add API endpoints for managing and running local scheduled scans
This change introduces a new set of Next.js API routes under \/api/scheduled-jobs\ to support local scheduled scans. Users can now create new scheduled jobs via a POST request, list existing jobs via GET, and update or delete specific jobs via PATCH and DELETE on \/api/scheduled-jobs/\[id\]\. Additionally, a new endpoint at \/api/scheduled-jobs/\[id\]/run\ allows users to manually trigger a scheduled scan, which executes a Node.js runner script with a 76-minute timeout and returns detailed results including any roles found or error messages if the scan fails.
web/src/app/api/scheduled-jobs · high confidence
Add Apify plugin for running actors as a keyed job source
Users can now enable the Apify provider plugin to run pre-built Apify actors and map their dataset items into the scanner's job shape. This is a keyed provider that only activates when a \portals.yml\ entry explicitly sets \provider: apify\, requiring an \APIFY\_TOKEN\ in the environment. The plugin handles the actor execution lifecycle (start, poll, fetch) with robust retry and timeout logic, normalizes actor IDs, and safely caches job descriptions locally under \jds/\ when a \field\_map.description\ is configured.
plugins/apify · high confidence
Add CV Studio page for reviewing and customizing tailored CVs
A new CV Studio page has been added at the /pipeline/\[id\]/cv-studio route, allowing users to review and customize a tailored CV. The implementation includes a loading state with skeleton placeholders and a dynamic page component that renders the CvStudio interface for a specific pipeline ID.
web/src/app/pipeline/\[id\]/cv-studio · high confidence
Add Catppuccin Latte light theme with auto-detection
The dashboard now supports the Catppuccin Latte light theme alongside the existing Mocha dark theme. Users can explicitly select the light theme via the 'catppuccin-latte' flag, or use 'auto' (or leave it empty) to let the dashboard automatically choose between Mocha and Latte based on the terminal's background color. Additionally, a new test suite ensures that all defined themes include every required color field, preventing unstyled UI elements.
dashboard/internal/theme · high confidence
Add Notion plugin to mirror application tracker
Introduces a new opt-in Notion plugin that mirrors the local application tracker to a Notion database and allows searching for job leads. The plugin exports tracker rows (company, role, status, score) to a user-specified Notion database and can search for records with job URLs. It includes a fix to correctly parse slash-formatted scores (e.g., '4.2/5') to prevent data mangling.
plugins/notion · high confidence
Add Ukrainian language and market support
The application now includes full Ukrainian language support, adding dedicated market modes and documentation to assist users targeting the Ukrainian job market. Also in this area: Optional years-of-experience filter for job postings; New role-clarity check, interview clarification drafts, and PDF performance improvements; Claude Code plugin support and updated documentation visuals; Fixes for CV template rendering, deduplication, and liveness checks; Agent inbox resolves by stable file position and validates targets; and 6 more.
(repo-wide) · high confidence
Add funding manifest URLs to .well-known
A new .well-known/funding-manifest-urls file has been added to the repository, listing five URLs that point to funding.json files hosted on GitHub (both in the career-ops-hq and santifer repositories). This change enables automated tools and users to discover the project's funding sources via the standard .well-known funding manifest protocol.
.well-known · high confidence
Batch runner gains multi-CLI support, company scoping, and JD rescue
The batch runner now supports multiple agent CLIs (opencode, gemini, qwen, grok) alongside the default claude worker, with parallel execution restricted to claude. A new --company flag filters offers by employer name, and a JD rescue module recovers job descriptions from structured data when visible text is thin. The runner also introduces --min-score to skip low-scoring offers, --skip-pdf to omit PDF generation, --worker-timeout to limit execution time, and --thin-action to control behavior when JD fetching fails.
batch · high confidence
Dashboard UI: new analytics screens, pipeline enhancements, and robustness fixes
The dashboard now includes a Progress analytics screen (funnel, rates, weekly activity) and a Stats analytics screen (archetype, work mode, location, pay, score/seniority charts, and insights). The pipeline view adds sortable Location, Pay, Last-contact, and Posted columns, a /-key live search, a RESPONDED filter tab, rejected/discarded tabs, tracker IDs, a POSTED column, and a date column that shows the latest status change. PDF handling now resolves the index from the tracker workspace, excludes cover letters from CV manifest lookups, and keeps artifacts in the workspace with symlinks. The report viewer supports clickable http(s) links, aligned tables, width-aware markdown rendering, and an in-viewer status picker. Several fixes improve correctness: unevaluated score rows render as the sentinel instead of 0.0, 0.0% rates are hidden when no applications exist, dates are reported in calendar days, open failures surface in the UI, control characters in the help-bar flash are neutralized, and status lifecycle side effects are preserved. Tests were added to cover these behaviors.
dashboard/internal/ui · medium confidence
Dashboard gains light theme, analytics screens, and improved error handling
The standalone Go TUI now supports a Catppuccin Latte light theme with automatic detection based on the terminal background, alongside new Progress and Stats analytics screens for viewing application metrics. The interface includes a manual refresh shortcut, a discard reason picker with a learning loop, and a celebration flow for hired statuses. Additionally, the dashboard now correctly surfaces open-command failures (such as opening URLs or cover letters) as visible UI messages instead of writing to stderr, and supports internationalization via a language toggle.
dashboard · high confidence
Declarative HTML section partials for CV templates
The templates/sections directory now uses declarative HTML partials (e.g., awards.html, certifications.html, competencies.html, education.html, experience.html, projects.html, skills.html) that render CV sections via field placeholders (such as TITLE, ORG, YEAR, LOCATION, CONTEXT, BULLETS) and conditional blocks (e.g., ORG\_BLOCK, LOCATION\_BLOCK, CONTEXT\_BLOCK). This enables users to customize the HTML structure, class names, and field order within each section’s ENTRY zone without modifying core logic, and supports optional fields like Awards/Honors, Certifications, Core Competencies, Education location, and Work Experience context.
templates/sections · high confidence
Enhanced job application model with analytics metrics and derived fields
The application data model now includes new fields for tracking derived information such as location, work mode, pay range, and posting date, enabling more detailed sorting and filtering in the pipeline view. Additionally, new metric structures (ProgressMetrics and StatsMetrics) have been added to support analytics screens, including funnel stages, score distributions, weekly activity, and conversion rates, while fixing a display issue where zero rates were incorrectly shown as red before any applications were submitted.
dashboard/internal/model · high confidence
Experimental local-first web UI for career-ops
An opt-in, alpha-status web dashboard is now available in the web/ directory, providing a local-first view over the existing CLI data files (pipeline, applications, reports) without introducing a separate database or engine. The UI includes a Pipeline tracker, Explore scanner, and Apply form (which never auto-submits), along with Today, Analytics, and Config pages. The Analytics section now features a Pipeline Sankey chart to visualize application status flows. The local API is secured by a same-origin and loopback guard by default, with optional allowlists for hosts and origins via environment variables. The setup requires Node 22.6+ and uses Turbopack, with a specific configuration to pin the Turbopack root to prevent Windows postcss OOM issues. Also in this area: Focused job evaluations and accessibility improvements; Fixes for CLI config persistence, Node engine requirements, error parsing, and legitimacy tone detection; New local-first web UI with agentic assistant and application automation; Dashboard binds to loopback by default and supports inline status editing.
web · high confidence
Interview session transcript storage and format specification
The interview-prep module now includes a defined structure for storing machine-readable interview transcripts. A new README in the sessions directory specifies that transcripts are saved as individual Markdown files containing speaker labels, metadata (such as company, role, round type, and source), and optional competency tags. This change establishes the data format for outputs from 'debrief' and 'practice' modes, ensuring downstream analysis tools can parse interview history consistently while keeping actual session content gitignored for privacy.
interview-prep · high confidence
Introduce AI-powered web search and enhanced Explore UI
Adds an experimental AI search mode to the Explore page, allowing users to describe a role in plain language to hunt the open web for matches, complete with a live reasoning trace, cost tracking, and a mode toggle. Introduces new UI components including a natural-language search box, animated discovery states with source progress chips, and polished discovery cards with company logos and fit hints. The Explore provider now persists results across mode toggles and reloads, and the filter builder supports city/location keywords and improved paste handling.
web/src/components/explore · high confidence
Introduce CV Studio for editing tailored CVs
The CV Studio now allows users to review and customize a tailored CV by editing individual content blocks. This feature parses the tailored HTML into a structured tree of sections and bullets, compares them against the base CV to track changes, and enables users to rewrite specific bullets using their local AI CLI, hide or reorder items, and render the final edited version as a PDF.
web/src/lib/cv-studio · high confidence
Introduce CV Studio for reviewing and customizing tailored CVs
A new CV Studio interface is available for reviewing and customizing a tailored CV before applying. The studio provides a live preview of the document, a keyword coverage meter to track alignment with job requirements, and an editor for section and bullet-level changes (including reordering, hiding, and rewriting). Users can adjust length and tone presets, and save their final version for application.
web/src/components/cv-studio · high confidence
Introduce Inbox Triage with free filtering and cost-gated scoring
The inbox now features a dedicated Triage surface that lets you filter incoming job postings using free, client-side facets (freshness, source, seniority, location, and keywords) without consuming tokens. Unscored rows display an honest 'not scored' status rather than a fabricated match percentage. You can save jobs to a persistent shortlist tray, skip postings with an undoable action, and trigger an AI score on the shortlist only after seeing and confirming the explicit token cost estimate.
web/src/components/inbox · high confidence
Introduce local-first assistant API with persistent chat history and gated profile management
The web application now supports a local-first assistant experience by introducing new API endpoints that manage persistent conversation history and user profile data. The new \/api/assistant/chats\ endpoint allows users to list, read, save, and delete their conversation history, ensuring context is preserved across sessions. The main \/api/assistant\ route now integrates with this history and uses a 'propose-then-confirm' model for profile updates, where the assistant can suggest changes to the user's personalization file (\modes/\_profile.md\) via the new \/api/personalization\ endpoint, but actual writes require explicit user confirmation. Additionally, CV ingestion is handled locally via \/api/cv/ingest\, which proposes a markdown representation of the user's CV for review before it is saved via \/api/cv\. These changes shift the assistant from a stateless tool to one that remembers user context and handles sensitive data (CVs, personalization) with explicit user consent and local storage.
web/src/app/api/assistant · high confidence
Introduce local-first pipeline management and offer ingestion
This change adds the core infrastructure for a local-first web UI, enabling users to manage their job application pipeline directly from the browser. It introduces two new API routes: one to ingest discovered offers into the local pipeline and scan history, and another to expose the current pipeline state (inbox and applications) to the client. A new React context provider (\PipelineProvider\) is added to keep the UI in sync with the local data, automatically refetching on window focus to ensure the snapshot remains fresh after status updates or new scans.
web/src/app/api/explore/add, web/src/app/api/pipeline, web/src/components/pipeline · high confidence
Introduce local-first web UI with new core application views
This change introduces an experimental, opt-in local-first web UI (v0) that replaces the previous interface with a new application shell and a suite of dedicated views. The new shell provides a persistent sidebar navigation, a global assistant console, and a 'Back to Top' control. It includes a new Analytics view for tracking search progress, conversion rates, and role insights, alongside a new Apply view that acts as a form-proxy to pre-fill and review employer applications. The UI also features a Config form for managing local AI engine connections, an Assistant Console for conversational guidance, and report enhancements like company evidence cards and resolved company logos. Also in this area: Improve text color accessibility with a dedicated brand-text token.
web/src/components, web/src/components/ui · high confidence
Introduce opt-in plugin system with integrity pinning and SSRF guards
Users can now extend career-ops with external integrations (e.g., job boards, email ingestion, notifications) by placing plugin directories in \plugins/\ (bundled) or \plugins.local/\ (user-managed). This feature is opt-in: plugins only load when explicitly enabled in \config/plugins.yml\ and their required environment keys are present in \.env\. The system includes a curated registry for community plugins, integrity pinning via \plugins.lock\ to detect tampering or drift, and strict egress validation (\ctx.fetch\) to prevent SSRF attacks by restricting network access to declared \allowedHosts\. Plugins are isolated by design, with a 'human-in-the-loop' requirement preventing auto-submission, and all data writes are handled canonically by the engine to maintain data integrity.
plugins · high confidence
Introduce streamed AI drive API for application forms
Added a new POST endpoint at /api/apply/drive that allows the AI agent to drive a headed browser session to reach or fill an application form. The API accepts a session ID, CLI ID, and goal (reach or full), then streams real-time step updates via an NDJSON stream. It handles session validation, form detection, and finalization, ensuring the stream is properly closed even on errors or success, and returns specific error codes (400 for bad JSON, 404 for missing session) instead of generic 500s.
web/src/app/api/apply/drive · high confidence
Introduce usage meter API endpoint
A new API route at /api/usage has been added to expose Claude Code's per-message token and message counts. This endpoint aggregates usage across all projects in the local \~/.claude directory, calculating totals for 5-hour and 7-day rolling windows to align with account rate-limit scopes. It utilizes a background scanner that tracks file positions to read only appended transcript bytes, ensuring efficient performance with a 60-second response cache.
web/src/app/api/usage · high confidence
Introduces AI-driven form interpretation and agentic drive loop for job applications
The Apply module now includes an agentic fallback that uses a local LLM to interpret live job-application forms when deterministic extraction fails. This new capability captures interactive controls, sends them to a planner CLI for classification, and then executes actions (click, type, select) via a Playwright-based drive loop. The implementation enforces a strict 'never-submit' policy to prevent accidental application submission, handles navigation obstacles like cookie banners and multi-step forms, and ensures tailored CVs and cover letters are correctly matched to specific applications rather than the company's newest file.
web/src/lib/apply · high confidence
Introduces a unified action registry for assistant and UI interactions
Adds a new \registry.ts\ module that serves as the single source of truth for all assistant intents and UI actions. This change ensures that both the human interface and the agentic assistant dispatch the same actions, preventing drift between the two. The registry includes validation logic for arguments, gates side-effects, and handles specific fixes such as normalizing company keys for explore and registry operations, as well as handling application number padding.
web/src/app/actions · high confidence
Introduces a visual backdrop and state management for the Apply page
Adds a full-viewport 'magic processing' wallpaper that displays a blurred, enlarged screenshot of the application form behind the interactive content, with animated glow halos and dot grids that intensify during loading states. This visual layer is managed by a new \ApplyProvider\ context and an \ApplyBackdropMount\ component, which controls the backdrop's visibility and intensity based on the session status (idle, opening, prefilling, filling, ready, error).
web/src/components/apply · high confidence
Local document intake scaffold for profile updates
A new \documents/\ directory structure has been introduced to support local profile intake. This includes a \.gitkeep\ file to maintain the folder in version control and a \README.md\ that documents the expected subdirectories (\cv/\, \linkedin/\, \diplomas/\, \references/\) and explains how users can drop source documents there for the agent to process. The documentation clarifies that extraction is local, idempotent, and requires explicit user confirmation before any changes are made to profile files like \config/profile.yml\ or \cv.md\.
documents · high confidence
Local scheduled scan installation scripts for macOS and Linux
Added PowerShell and shell scripts to install and uninstall a local scheduled job that runs the scan runner every 15 minutes. On Windows, the script registers a Windows Scheduled Task; on macOS, it installs a launchd LaunchAgent; and on Linux, it adds a crontab entry. These scripts ensure the recurring scan worker runs under the current user's context without requiring root privileges.
web/scripts · high confidence
New 4 Day Week provider and shared infrastructure hardening
Added a new zero-auth provider for the 4 Day Week job board, which scrapes its public JSON API to aggregate reduced-hours roles. This release also introduces several shared infrastructure modules to the providers directory: a centralized HTML entity decoder (\_html-entities.mjs) to prevent code-point injection and ensure consistent decoding across all providers; a shared HTML-to-text pipeline (\_html-to-text.mjs) for safely stripping markup from job descriptions; an in-process DNS cache (\_dns-cache.mjs) to rate-limit resolver lookups and prevent DNS rebinding attacks; and an IP address guard (\_ip-guard.mjs) to block connections to private or reserved IP ranges. These changes collectively improve security posture and parsing reliability for all job-source providers.
providers · high confidence
New AI-powered job search API endpoint
Added a new server-side API route at \/api/explore/ai\ that orchestrates headless execution of user-configured AI CLIs (such as Codex or Claude) to perform job searches. The endpoint validates JSON input, resolves the specified CLI, and enforces strict read-only capabilities (WebSearch only) via a fencing mechanism to prevent unauthorized file writes or tool usage. It supports isolated execution environments for Codex, handles deduplication against known offers, and returns structured job candidate data.
web/src/app/api/explore/ai · high confidence
New AI-powered prefill endpoint for application forms
A new API route at \/api/apply/prefill\ has been introduced to automatically draft answers for application form fields using an external CLI planner. This endpoint accepts a session ID, optional company/application identifiers, and a CLI ID, then streams NDJSON diagnostics and the final JSON-parsed answers back to the client. It resolves the user's tailored CV from the current session, constructs a prompt based on the form fields and career ops memory, and executes the planner with strict input validation (returning 400 for malformed JSON or non-string identifiers) and idempotent error handling to ensure reliable prefilling.
web/src/app/api/apply/prefill · high confidence
New API endpoint for safe, atomic profile updates
A new API route at /api/profile has been introduced to handle user profile modifications. This endpoint validates incoming JSON payloads before writing, ensuring that existing profile.yml files are not overwritten if they contain invalid YAML structures or non-mapping roots, thereby preventing data loss. It performs deep-merges of proposed changes onto the existing configuration and utilizes atomic writes with backups to maintain file integrity. Additionally, the implementation attempts to preserve existing YAML comments during updates, falling back to a standard dump only when necessary, and seeds new profiles from an example file if none exist.
web/src/app/api/profile · high confidence
New API endpoint to detect local CLIs and remember the active agent CLI
A new API route at /api/clis has been added to support the local-first web UI. This endpoint detects which agnostic CLIs are installed on the user's machine and returns the currently remembered CLI for the career-ops agent, enabling the web interface to delegate tasks to local tools without requiring an API key.
web/src/app/api/clis · high confidence
New API endpoint to manage follow-up cadence settings
A new API route at /api/followups/cadence allows users to read and update follow-up cadence configurations directly in their profile.yml. The GET endpoint retrieves current overrides and derives pure defaults from the core system, ensuring the UI reflects accurate baseline values rather than stale copies. The POST endpoint enables users to adjust specific cadence knobs and toggle follow-ups on or off, using atomic writes that preserve existing profile content and comments while validating input to return 400 errors for malformed data instead of 500s.
web/src/app/api/followups/cadence · high confidence
New API endpoint to persist inbox skip actions to pipeline.md
A new API route at /api/inbox/skip has been added to allow users to persist skip/undo actions directly to the data/pipeline.md file. This endpoint accepts a JSON body with a URL and a boolean 'done' flag, and uses the career-ops scripts to flip the corresponding checklist row (toggling between unchecked and checked states). It returns specific HTTP status codes (400, 404, 409, 503) for various error conditions such as invalid input, unmatched URLs, or concurrent write conflicts, ensuring robust handling of inbox state changes.
web/src/app/api/inbox · high confidence
New API endpoint to pre-fill application forms using attached CVs
A new API route at /api/apply/fill has been introduced to automatically populate application forms with data extracted from a CV attached to the specific application. This endpoint accepts a session ID, form fields, and optional company/application identifiers to resolve the correct tailored CV server-side, ensuring the pre-filled answers correspond to the current application rather than the company's most recent CV. It also includes input validation to return 400 errors for malformed JSON or missing session IDs, improving reliability for users initiating the application process.
web/src/app/api/apply/fill · high confidence
New API endpoint to serve tailored CV and cover letter PDFs
A new API route at \/api/cv-pdf\ has been added to serve tailored CV and cover letter PDFs generated by the career-ops tool. This endpoint supports three resolution methods: by a specific report number (\n\), by an application tracker row (\application\), or by company slug (fallback). It ensures that cover letters are only served when explicitly indexed by report number, prevents non-ASCII filenames from breaking access, and handles cases where a company lacks a usable slug by returning a 404 instead of serving an unrelated file.
web/src/app/api/cv-pdf · high confidence
New API endpoints for CV Studio bullet rewriting and editing
The CV Studio now supports rewriting individual resume bullets using an AI CLI and saving manual edits. A new POST endpoint at \/api/cv-studio/\[n\]/rewrite\ allows users to provide an instruction and a CLI ID to regenerate a specific bullet, with validation to ensure the AI-generated text does not introduce unsupported claims. Additionally, the main studio endpoint at \/api/cv-studio/\[n\]\ now handles PUT requests for autosaving manual edits and POST requests for finalizing changes, which triggers the regeneration of the edited PDF. These changes enable users to tailor their CV content more precisely through both AI assistance and direct editing.
web/src/app/api/cv-studio · high confidence
New API endpoints for closing apply sessions and saving run logs
Added two new API routes to support the experimental local-first web UI. The \/api/apply/close\ endpoint allows the client to explicitly close an apply session, freeing the off-screen browser tab immediately rather than waiting for a background prune. The \/api/runs/save\ endpoint persists finished worker logs as markdown files in a local \.career-ops-web/runs\ directory, enabling the CLI assistant to read past run history.
web/src/app/api/apply/close, web/src/app/api/runs · high confidence
New API endpoints for follow-up seeding and PDF readiness
Two new API routes have been added to support the apply checklist workflow. The \/api/followups/seed\ endpoint allows seeding the first follow-up pin by accepting a tracker row number, while the \/api/pdf-ready\ endpoint marks a report's PDF as ready by accepting a report number and invoking the underlying core script to update the tracker status.
web/src/app/api/followups/seed, web/src/app/api/pdf-ready · high confidence
New API endpoints for managing and verifying portal configurations
Added two new API routes under \/api/portals\ to handle portal configuration and health checks. The \POST /api/portals\ endpoint allows users to update role and location filters in \portals.yml\, preserving existing comments and structure while validating input (returning 400 for invalid JSON). The \GET /api/portals/verify\ endpoint orchestrates the core \verify-portals\ script to check the status of configured companies (live, empty, broken, or skipped), providing detailed feedback on portal health to prevent silent failures in job scanning.
web/src/app/api/portals · high confidence
New ATS-friendly single-column CV template pack
A new 'ATS Friendly' template pack has been added to the templates/ats directory, providing a single-column layout optimized for Applicant Tracking Systems. The main template (cv-template.ats.html) establishes the base structure with specific CSS variables for page margins and font settings, including an opt-in style token to prevent job entries from splitting across pages. The pack includes dedicated section partials for Work Experience, Education, Projects, Skills, Core Competencies, Certifications, and Awards, all designed to use simple, linear HTML structures (avoiding tables or complex flex layouts) to ensure parsers can read content in a logical, left-to-right, top-to-bottom order.
templates/ats · high confidence
New Analytics dashboard with Progress, Insights, and Search Stats views
The analytics section now features a dedicated dashboard page that presents data across three distinct views: Progress, Insights, and Search Stats. Users can view their application progress through a Pipeline Sankey diagram and funnel rates, while the Insights view provides search-specific metrics and generated insights. The implementation includes a loading state skeleton for improved perceived performance and resolves the active view based on URL parameters, ensuring that incomplete data in one area (like the core progress install) does not break the other views.
web/src/app/analytics · high confidence
New CV templates, CJK support, and agency licensing data
The templates area now includes five new named CV templates (Compact, Executive, Jake, Leadership, Modern) alongside a new CJK LaTeX variant that uses the tectonic engine for Chinese, Japanese, and Korean typography. The system also introduces a jurisdiction table for agency/recruiter licensing regimes (templates/agency-licensing.yml) used by the AI screening mode, a new market calibration benchmarks file (templates/benchmarks.yml), and a user-managed company blacklist (templates/blacklist.example.md) that gates scanning and application modes.
templates · high confidence
New CostBadge component for displaying cost semantics
A new CostBadge component has been added to the web UI to visually indicate cost status using specific color and icon semantics: green with a leaf icon for free/positive actions, and a muted neutral tone with a coin icon for spend (token usage). This component ensures that spend indicators do not conflict with primary action colors (orange) and maintains accessible contrast across both light and dark themes.
web/src/components/cost · high confidence
New Documents dashboard for browsing generated CVs and cover letters
A new Documents page has been added to the web application, allowing users to browse a grid of their generated CVs and cover letters. The page fetches document metadata via a new API endpoint at /api/documents and displays each item with its filename, type (CV or Cover Letter), date, and format. Users can open the corresponding PDF files directly from the dashboard. A loading state and error handling are included to manage the data retrieval process.
web/src/app/api/documents, web/src/app/documents · high confidence
New Follow-up Tracker with configurable cadence and status controls
The Follow-ups page now includes a full tracker that lets you manage and monitor your job applications. You can configure the follow-up cadence (e.g., days between nudges, max follow-ups) via a new settings panel, which saves to your profile and syncs with the CLI. The main view displays a sortable, filterable list of applications with urgency levels, allowing you to log follow-up details (date, channel, notes), update application status, and pin custom next-follow-up dates. It also highlights companies that have gone silent past a courtesy threshold.
web/src/components/followups · high confidence
New Gmail plugin for ingesting job leads
A new bundled Gmail plugin has been added to pull job leads from a configured Gmail label into the pipeline. It uses OAuth credentials to fetch emails, extracts clean job URLs from the body, and parses role/company information from subject lines. The plugin includes security measures such as validating DMARC authentication via Gmail's own headers and filtering out page assets (like logos) and tracking URLs to prevent noise. It maintains a processed-message cursor to avoid re-ingesting old emails and stores its state under the data root.
plugins/gmail · high confidence
New H-1B Sponsor Check plugin for verifying employer sponsorship history
Adds the \h1b-sponsor\ plugin, which allows career-ops users to verify whether a US employer has a history of filing H-1B and PERM applications with the Department of Labor. The plugin resolves a company name to a specific DOL entity and returns a sponsorship tier (strong, moderate, staffing-shop, weak, none) based on filing volume, recency, and green card evidence. Lookups run locally by default using a downloadable DOL index (staying on the machine), with an opt-in HTTP backend available via the \H1B\_API\_BASE\ environment variable. The plugin includes a CLI (\check.mjs\) for manual checks, a companion agent skill for automated evaluations, and a dedicated installer script to manage the local data index.
plugins/h1b-sponsor · high confidence
New analytics views for conversion funnels, pipeline flows, and search insights
The Analytics section now includes three new client-rendered components that surface core analytics data directly to the user. The Funnel Rates view displays cumulative conversion stages and rates fetched from /api/stats, including handling for empty trackers and small sample sizes. The Pipeline Sankey view visualizes application status transitions as a flow diagram using data from the tracker's status log, with an accessible screen-reader table for non-visual users. The Search Insights view presents pattern analysis from /api/patterns, showing recommended thresholds, breakdowns by dimension, and vendor-specific advance rates, while clearly indicating when data is insufficient to make claims. Also in this area: Fixes for Windows stability, logo caching, analytics display, and inbox reconciliation.
web/src/components/analytics · high confidence
New apply checklist page for job applications
A new apply checklist page has been added for each application, providing users with a consolidated view of their application status, including resume and cover letter PDF readiness, copied answers, and relevant links. The page includes dedicated error and loading states to handle cases where the application data or reports cannot be loaded, ensuring a smooth user experience even when underlying data is missing or unavailable.
web/src/app/pipeline/\[id\]/apply · high confidence
New diagnostic endpoint for bug-report data shapes and system state
A new API route at /api/report/shape has been added to provide structural diagnostics for the in-app bug reporter. This endpoint returns a JSON summary of the application's environment (Node version, platform, architecture), setup status (prerequisite file presence, data availability), and data-shape fingerprints (counts of candidate vs. parsed rows in inbox and tracker files, number of reports and PDFs). It also exposes capability probes such as JSON scan support and tracker deletion ability. This allows users to include precise, privacy-preserving context (e.g., "12 candidate rows, 5 parsed") in bug reports without exposing sensitive file contents.
web/src/app/api/report · high confidence
New example files and documentation for career-ops data formats
Added reference examples and documentation in the \examples/\ directory to demonstrate expected data structures and conventions. This includes \cv-example.md\ and \resume-example.md\ for standard CV/resume formatting, \article-digest-example.md\ and \sample-report.md\ for evaluation pipeline outputs, and \ats-normalization-test.md\ as a regression fixture for Unicode normalization. A new \dual-track-engineer-instructor/\ folder provides a complete profile configuration (\profile.yml\) and CV (\cv.md\) for candidates with hybrid engineering and teaching roles, including guidance on configuring multiple primary archetypes and compensation ranges. Additionally, \latex-tex/\ fixtures support the new LaTeX extraction capabilities.
examples · high confidence
New explore API endpoint with China employer discovery preset
A new API route at /api/explore has been added to handle job discovery requests, introducing an opt-in preset for discovering employers in China. The endpoint validates the request body, returning a 400 status for unknown presets or invalid source configurations, and streams discovery results as NDJSON. It supports both standard ATS scanning and the new employer-specific discovery flow, ensuring proper error handling for missing scanner scripts.
web/src/app/api/explore · high confidence
New follow-up logging API with input validation and atomic writes
The web application now exposes a new API endpoint at /api/followups/log to record and remove follow-up entries. This endpoint validates all incoming JSON bodies, rejecting malformed IDs, invalid dates, and missing required fields with 400 errors instead of server crashes. It ensures data integrity by using atomic writes and locking mechanisms to prevent race conditions with the core seeder, and supports both adding new follow-ups (POST) and deleting specific entries by ID (DELETE).
web/src/app/api/followups/log · high confidence
New home dashboard with decision cards, follow-up tracking, and onboarding progress
The home page now features a new 'Today' dashboard that consolidates job matches, follow-ups, and setup progress into a single view. Users can take immediate action on new matches via Decision Cards, which display AI-generated reasons and allow them to mark applications as 'Applied' or 'Skip' directly from the home screen. A Follow-Up Tracker section surfaces due follow-ups with one-tap logging and snooze capabilities. Additionally, an Onboarding Progress panel guides new users through profile setup prerequisites, while a First-Run Home experience provides a streamlined entry point for CV upload without requiring an account.
web/src/components/home · high confidence
New local-first CV ingest and parsing workflow
Introduces a new CV ingestion component that allows users to paste text or drop files (Markdown, TXT, PDF, DOCX) for parsing. The component supports a fast path for plain-text files without requiring an external CLI, while PDF/DOCX files are sent to a local AI CLI for extraction. After parsing, users review the extracted content and save it, which automatically redirects them to the Job Explorer with pre-filled filters and triggers an initial job scan. Also in this area: Improve text color accessibility with a dedicated brand-text token; Focused job evaluations and accessibility improvements.
web/src/components/cv · high confidence
New local-first web UI with agentic assistant and application automation
This change introduces the experimental local-first web UI (opt-in alpha), establishing the core application structure and backend routes. It adds a centralized action registry to synchronize UI controls with the AI assistant, enabling the assistant to perform actions like navigation, pipeline filtering, and job evaluation. The update includes new API routes for the assistant chat (with persistent history and conversation context), application session management (opening, filling, and driving forms via a headed browser), and document generation (serving tailored CVs and cover letters as PDFs). Additionally, it implements an Analytics page with progress tracking, funnel rates, and search statistics, alongside a company evidence card to provide deeper insights into application history. Also in this area: Fixes for Windows stability, logo caching, analytics display, and inbox reconciliation.
(repo-wide) · high confidence
New memory API endpoint for user facts
A new API route at /api/memory has been added to allow the web UI to store durable facts about the user. The GET endpoint retrieves current memory, while the POST endpoint accepts a JSON object containing a 'fact' string and persists it to the canonical modes/\_profile.md file, ensuring consistency with CLI/TUI tools. The endpoint validates that the request body is a valid JSON object and that the fact is non-empty, returning a 400 error for invalid inputs. If the profile file cannot be read or written, it returns a 500 error with a specific message indicating the file path and that the fact was not saved.
web/src/app/api/memory · high confidence
New npx career-ops scaffolder for one-command workspace setup
Users can now run \npx career-ops init\ to automatically clone the latest release of the career-ops repository and install dependencies in a single step. The new \scaffolder/bin/cli.mjs\ script detects installed AI coding tools (such as Claude Code, Codex, Grok, and Pi) to tailor the post-installation instructions. It also includes a Windows-specific fix in \npm-command.mjs\ to reliably run \npm install\ without shell-injection risks, and \skill-entrypoints.mjs\ ensures that AI skill pointers are correctly materialized on filesystems that do not support symlinks.
scaffolder/bin · high confidence
New opt-in plugin system with governance and registry support
Users can now extend the platform using an opt-in plugin system that includes an engine, supply-chain governance, and a registry. The \plugins/\_template\ directory provides a starter kit for community plugins, defining the structure for manifests, hooks (such as ingest, export, and notify), and configuration via \config/plugins.yml\ and \.env\ secrets. The system enforces security by restricting network egress to approved hosts via \ctx.fetch\ methods and requires human-in-the-loop consent for plugin activation. A smoke test ensures plugins adhere to the declared hooks and manifest structure.
_plugins/\template · high confidence
New scripts for syntax checking, ATS export, and local web UI
This update adds several new scripts to the repository. A new \check-syntax.mjs\ script provides a zero-dependency syntax linter for all \.mjs\ files, running in a single process to improve performance. The \export-ats-text.mjs\ script introduces a new feature allowing users to export their profile and experience data as plain text formatted for copy-pasting into Applicant Tracking System (ATS) forms. Additionally, the \ui-server.mjs\ script adds a local web UI launcher that starts a Next.js development server, handling port/host configuration and automatically opening the browser. Finally, the \followup-sweep.sh\ script enables unattended follow-up sweeps via launchd or cron, generating draft emails and LinkedIn messages for overdue items without sending them.
scripts · high confidence
New token usage tracking and cost estimation utility
A new \utils/token-tracker.mjs\ module has been added to centralize token counting and cost calculation for AI model interactions. This utility provides a \TokenAccumulator\ class to track token usage across pipeline steps, a \normalizeOpenAIUsage\ function to standardize usage data, and an \estimateCost\ function that applies pricing rates for models including MiniMax, OpenAI, Gemini, DeepSeek, and Anthropic. It also includes a \formatBreakdown\ function to generate human-readable reports of token consumption and estimated costs per step.
utils · high confidence
New web library modules for analytics, assistant chat, and security fencing
The web application now includes a suite of new core library modules in \web/src/lib\ that power several key capabilities. \analytics-data.ts\ and \analytics-metrics.mjs\ provide the logic for the new Analytics dashboard, computing progress funnels, score distributions, and enriching application data with archetype, location, and pay details. \assistant-chat-store.mjs\ and \assistant-history.mjs\ implement the persistent storage and context management for the new assistant conversations, including message cleaning and title generation. Security and runtime control are enhanced by \claude-invocation.mjs\, which defines tool scopes and fencing for AI agents, and \cli-fencing-probe.mjs\, which probes CLI binaries for supported flags to ensure safe execution. Additionally, \act-envelope.mjs\ handles parsing of assistant action envelopes, \clean-chips.mjs\ manages keyword input with a new 512-chip cap, and \bind-host.mjs\ enforces loopback binding and host validation for the dashboard.
web/src/lib · high confidence
Seed startup discovery from VC portfolios (Y Combinator, a16z)
Added a new seeding mechanism that pulls public company lists from Y Combinator and Andreessen Horowitz (a16z) to instantly cover hundreds of VC-backed startups, feeding them into the existing ATS job discovery pipeline. This provides a high-signal starting point for finding startup roles without waiting for companies to appear in public ATS directories, with support for combining these seeds alongside regular ATS sources.
seeds · high confidence
Spanish language support added to the dashboard TUI
The dashboard's internalization system now includes a Spanish (es) locale, making it available for users via the language toggle key. This change ensures that Spanish-speaking users are no longer stranded in English and that the language cycling mechanism correctly includes Spanish alongside English and Turkish. The update also introduces tests to guarantee that all UI strings are present in every supported language and that the language toggle can reach all available locales.
dashboard/internal/i18n · high confidence
Behavioural changes
Added loading skeletons for Config and CV pages
The web application now displays visual loading feedback while the Configuration and CV Editor pages are being fetched. New \loading.tsx\ files have been added to the \config\ and \cv\ route directories, utilizing skeleton components to outline the layout of the configuration form and the CV editor interface respectively, improving the perceived performance during data loading.
web/src/app/config, web/src/app/cv · high confidence
Application tracker state upgraded to v1.18 with 'via' column and salary observations
The test fixtures for the application tracker have been updated to reflect the v1.18 state, introducing two key data changes. First, the \applications.md\ tracker now includes a 'Via' column to record the source of each application (e.g., 'Hays'), replacing the previous structure. Second, a new \salary-observations.tsv\ file has been added to track salary data, such as target bands and actual offers. These changes are validated by the \expected.json\ file, which now asserts the presence of the 'via' column and the count of salary observations.
test-fixtures · high confidence
Enhanced bug-report diagnostics and deduplication via stable fingerprints
The in-app bug reporter now captures richer diagnostic data, including core version, API error status, and structural data-shape counts (without contents), while enforcing strict PII scrubbing of user descriptions and logs. A new client-side ring buffer captures recent errors and fetch failures, and a deterministic fingerprint (based on route, error class, and structural flags) is generated to enable click-gated deduplication of similar issues at write time. The report body format is now versioned (v1) to ensure stability for downstream triage tooling.
web/src/lib/report · high confidence
Follow-up Tracker API and Windows Console Fix
The application now exposes a dedicated /api/followups endpoint that retrieves follow-up data from the core's followup-cadence script, distinguishing between due items and upcoming ones based on urgency. This supports a new Follow-up Tracker page with logging, history, and cadence settings, while also ensuring that only actually due follow-ups are surfaced in the home card. Additionally, a new /api/doctor endpoint validates the core setup, and a fix prevents Windows users from encountering stale console errors that previously masked setup failures.
web/src/app/api/followups · high confidence
Improved CV upload error feedback and readiness hints
When a CV upload is refused by the backend, the system now displays the specific reason for the rejection (such as missing AI CLI configuration or a generic HTTP error) instead of a vague 'couldn't read your CV' message. Additionally, a new readiness check provides users with contextual hints—like suggesting they add experience or skills—if their CV content is too short or lacks key sections to generate a meaningful score, while still allowing them to save the document.
web/src/lib/cv · high confidence
Local dashboard API restricted to same-origin and loopback by default
A new proxy middleware in web/src/proxy.ts now gates all /api requests with a same-origin and loopback check, preventing cross-site and LAN-based access by default. Users can opt-in to allow specific hosts or origins via the CAREER\_OPS\_WEB\_ALLOWED\_HOSTS and CAREER\_OPS\_ALLOWED\_ORIGINS environment variables, which is necessary for local companion clients like browser extensions that originate from chrome-extension:// URLs.
web/src · high confidence
Pipeline view gains new tabs, sort modes, and i18n support
The pipeline screen now includes additional filter tabs for 'Assessment', 'Responded', 'Rejected', and 'Discarded' statuses, alongside new sort modes for Location, Pay, Last-contact, and Posted date. User-facing tab labels are now localized via the i18n module instead of being hardcoded in Spanish. The code also introduces message types for opening job URLs, generating/viewing CV PDFs, and handling status updates with discard reasons, supporting the broader dashboard features for these capabilities.
github.com/santifer/career-ops/dashboard · high confidence
Route-aware loading feedback and precise cover resolution for pipeline reports
The pipeline report view now displays a skeleton loading state while data is being fetched, improving perceived performance. Additionally, the logic for determining if a tailored cover letter is available has been corrected to resolve the cover specifically for the current application rather than defaulting to the company's most recent cover, ensuring users see the correct document for the application they are viewing.
web/src/app/pipeline/\[id\] · high confidence
Scheduled scans route now includes a loading state
The scheduled scans page now displays a skeleton loading state while content is being fetched, preventing the previous page from remaining visible during navigation. This change ensures a smoother user experience by providing immediate visual feedback when accessing the scheduled scans section.
web/src/app/scheduled-scans · high confidence
Status updates now delegate to the core CLI with improved error handling and locking
The /api/status endpoint has been rewritten to delegate status changes to the core CLI (set-status.mjs) instead of directly editing the tracker file. This change ensures that status updates are protected by the shared tracker lock, preventing race conditions with other writers. It also improves error handling by validating input earlier, excluding child process stderr from the response to avoid leaking server paths, and mapping CLI exit codes to appropriate HTTP status codes (e.g., 400 for bad requests, 503 for lock contention). Additionally, the route now supports undoing the last status change and detaches JD archive processes to prevent timeouts.
web/src/app/api/status · high confidence
Today page queue ordering and headline reliability improvements
The Today page now ranks scored-but-undecided applications by score (highest first) and date, replacing the previous unstable order that depended on raw data row sequence. This ensures the most relevant decisions appear at the top. Additionally, the page's headline and status indicators are more reliable: they now correctly handle partial data loads and errors by showing specific counts for available sources rather than hiding all work or displaying stale numbers, and they properly distinguish between loading, unavailable, and partial states.
web/src/lib/home · high confidence
Fixes
Added placeholder files for batch, JDS, and output directories
Added .gitkeep placeholder files to the batch/logs, jds, and output directories to ensure these directories are tracked by version control even when empty.
batch/logs, jds, output · high confidence
Beta bug reporter now signals when duplicate search fails
The beta bug reporter's duplicate-check feature now clearly distinguishes between a successful search that found no matches and a search that could not run at all. Previously, network errors or API failures (such as HTTP 422 from a repository move) were silently treated as 'no similar issues found,' hiding the error from users. The underlying search module now returns \null\ to indicate a failure state, allowing the UI to inform the user that the check could not be completed rather than falsely claiming no duplicates exist.
web/src/lib/beta · high confidence
Fix company matching and evaluation suppression in What's New API
The What's New API now correctly handles non-Latin characters in company names by deriving matching keys from the core text-key library instead of a previous ASCII-only implementation that caused collisions (e.g., 'Škoda' vs 'Koda'). Additionally, the API fixes a bug where evaluating a single role from an employer would hide the entire company's job board; matches are now suppressed based on the specific company-and-role combination rather than the company alone.
web/src/app/api/whats-new · high confidence
Fix cover PDF serving for non-ASCII filenames and specific applications
The new cover PDF API route now correctly serves cover letters for applications with non-ASCII filenames and ensures the system retrieves the tailored cover for the specific application rather than defaulting to the company's most recent one. This resolves previous issues where 'View cover' links and availability checks disagreed on file existence, and fixes 400 errors caused by invalid company slugs.
web/src/app/api/cover-pdf · high confidence
Fixes for CLI flag parsing, CV structure detection, and symlinked entrypoints
This update introduces several shared library modules to fix silent failures and improve robustness across the CLI and CV generation tools. The new \lib/cli-flags.mjs\ centralizes argument parsing, ensuring that \--flag=value\ syntax is correctly recognized and that unrecognized or mistyped flags fail fast instead of silently falling back to defaults. \lib/cv-markdown.mjs\ adds support for Pandoc-flavored Markdown, allowing CV checks to correctly parse sections and entries that use Pandoc's specific formatting (such as bracketed spans and smart typography) which were previously ignored. \lib/cv-payload-schema.mjs\ enforces a strict key contract for CV payloads, preventing silent data loss by rejecting payloads with unknown section keys or missing required fields. Additionally, \lib/is-main-module.mjs\ fixes a bug where CLI scripts accessed via symlinks would exit silently without running, by canonicalizing path comparisons. Other libraries like \lib/ascii-fold.mjs\ and \lib/context-budget.mjs\ provide foundational utilities for name normalization and token budget management used by these features. Also in this area: Fixes Playwright error parsing, optimizes history line parsing, and adds years-of-experience support; Support for MiniMax-M3.1-Flash model and new shared PDF painting service; Fix Windows CLI launching and detached process console behavior; Fixes for CV template warnings, scan history parsing, OpenAI headers, status hooks, and test concurrency.
lib · high confidence
Fixes for Windows stability, logo caching, analytics display, and inbox reconciliation
This update resolves several issues affecting reliability and data accuracy on Windows and in the analytics/inbox views. On Windows, child processes launched by the web API and core libraries now use \process.execPath\ and \windowsHide: true\ to prevent crashes caused by inherited console handles. The logo lookup logic has been refactored to distinguish between a definitive miss and a failed request (timeout, network error, or server error); previously, transient failures were cached as permanent misses, so the system now retries failed lookups. In the analytics breakdown, buckets with no submissions no longer incorrectly display a 0% conversion rate; instead, the rate is shown as absent. The quality-threshold recommendation panel now appears whenever the core provides reasoning, even if a specific threshold value is not recommended, ensuring users see the underlying sample-size context. Finally, the pending inbox now automatically marks postings as done if they have been evaluated via a direct URL, preventing stale entries from lingering. Also in this area: Analytics progress and insights views now display data with appropriate honesty signals.
(repo-wide) · high confidence
Improve text color accessibility with a dedicated brand-text token
The web interface now uses a new \brand-text\ CSS variable for all small text, badges, icons, and links, replacing the previous raw \brand\ color. This change addresses accessibility issues where the original brand color failed contrast requirements in light mode, ensuring better readability across the UI. Additionally, the Explore page now passes only the fields it matches on to the view, and the Pipeline page includes tracker shape data to better distinguish empty states from search results. Also in this area: New local-first web UI with agentic assistant and application automation.
web/src/app/pipeline/compare · high confidence
Improved error handling and URL validation in Apply session API
The Apply session API now provides more robust error handling for clients. It validates that the request body is a valid JSON object, returning a 400 Bad Request instead of a 500 error if the input is malformed. Additionally, it enforces URL safety by rejecting private or blocked URLs with a 400 response, and ensures that stale CLI identifiers are properly resolved or handled gracefully during session creation.
web/src/app/api/apply/session · high confidence
Introduce local job tracking and improved error visibility in the web UI
The web interface now tracks job execution state locally using a new \JobsProvider\ and \WorkerCard\ components, persisting job history in \localStorage\ so that interrupted jobs are correctly marked as errors on reload. This change validates the saved CLI ID against installed versions before starting a run to prevent failures caused by missing CLIs, and displays specific, cause-aware error hints (including fencing notices) instead of generic authentication prompts when jobs fail. Also in this area: Improve text color accessibility with a dedicated brand-text token; Focused job evaluations and accessibility improvements; Assistant conversations now persist to disk and CLI fallbacks are handled gracefully.
web/src/components/jobs · high confidence
Restored and enhanced the beta bug-reporting banner
The beta banner component has been restored and updated to provide a local-first, privacy-preserving way for users to report issues. It now includes a click-gated deduplication feature that searches for similar open GitHub issues before submission, and clearly informs users when this check cannot run. The report dialog pre-fills a GitHub issue with scrubbed diagnostic data, ensuring no sensitive personal information is sent until the user explicitly confirms.
web/src/components/beta · high confidence
Web core library mirrors core checkout logic for paths, locking, and text normalization
The web application now includes a comprehensive set of core utility modules that replicate the behavior of the user's local career-ops checkout, ensuring consistency between the web UI and the CLI. This includes \data-root.mjs\ and \code-root.mjs\ to resolve the correct data and script directories, \ascii-fold.mjs\ and \normalize-text-key.mjs\ for Unicode-safe company and role matching, and \followups-lock.ts\ to prevent race conditions when writing follow-up data. Additional modules handle CV template resolution, pipeline sorting, logo caching, and atomic file writes, all designed to prevent silent data loss or incorrect behavior when the web runtime is separated from the core checkout.
web/src/lib/core · high confidence
Test coverage
Add visual regression tests for CV templates and layout invariants; Added test coverage for Gmail plugin security fixes and H1B-Sponsor plugin contracts; Added tests for dashboard data layer logic and path resolution; Added tests for pipeline sort tie-breaking and score parsing; Expanded provider test coverage and shared utility validation; Expanded test coverage for core analysis and CLI validation logic; Expanded test coverage for web application logic and infrastructure; Expanded test fixtures for ATS providers, deduplication logic, and data safety; New test helper utilities for batch, spawn, and concurrency testing.
Dependencies
Major dependency upgrades and scaffolder introduction
This release upgrades core dependencies across the project: the web interface now uses Next.js 16.3.4, React 19.2.5, and js-yaml 5.0.0, while the root CLI updates Playwright to 1.63.0, js-yaml to 5.3.0, and adds the Google Generative AI SDK. The Go-based dashboard bumps the Go toolchain to 1.26.0 and updates terminal libraries (charmbracelet/x/ansi, golang.org/x/sys). Additionally, a new scaffolder package (@santifer/career-ops) is introduced to provide a one-command installer for the pipeline.
(dependencies) · high confidence
Housekeeping
Creation of empty data directories
Empty placeholder files (.gitkeep) were added to the data, data/offers, and data/parser-output directories to ensure these folders are tracked by version control.
data · high confidence
Written by watchdog.canine.dev from the codebase's own history, inside the signed delivery this page is composed from.
How this codebase got here
This is the PUBLIC form of this artifact. Findings are listed in full, but the details of SECURITY findings — which rule fired, in which file, on which line, and how to fix it — are deliberately withheld, and any secret-scanner results are excluded entirely. Where detail is absent here it was REMOVED FOR PUBLICATION; it is not missing from the analysis. The complete artifact is available from the repository owner.
Score
- CAI 52 → 53 (+0.1)
Lenses
- Code Health 44 → 45 (+1.4)
- Architecture 85 → 79 (-5.7)
- Maturity 68 → 82 (+13.2)
- Readiness 56 → 59 (+2.7)
- Security 77 → 77 (-0.1)
- Accessibility 57 → 50 (-6.9)
- Performance 62 → 67 (+4.7)
Resolved (113)
- Documentation: no contributor guidance (README.md)
- Documentation: no licence statement (README.md)
- Documentation: no project overview (README.md)
- FunctionTooLong: pipeline-lock.acquirePipelineLock (pipeline-lock.mjs)
- FunctionTooLong: tracker-utils.acquireTrackerLock (tracker-utils.mjs)
- High: security finding (details withheld)
- High: security finding (details withheld)
- High: security finding (details withheld)
- High: security finding (details withheld)
- Hotspot: pipeline-lock.mjs (pipeline-lock.mjs)
- Hotspot: portal-health-lock.mjs (portal-health-lock.mjs)
- Hotspot: tracker-utils.mjs (tracker-utils.mjs)
- Hotspot: web/src/app/api/logo/route.ts (web/src/app/api/logo/route.ts)
- Hotspot: web/src/lib/core/pipeline.ts (web/src/lib/core/pipeline.ts)
- Medium: security finding (details withheld)
- Medium: security finding (details withheld)
- Medium: security finding (details withheld)
- Medium: security finding (details withheld)
- Medium: security finding (details withheld)
- Medium: security finding (details withheld)
- …and 93 more
New (414)
- Dependency advisory scan runs only on code events
- FileTooLong: ./browser-extract.mjs (browser-extract.mjs)
- FileTooLong: ./followup-cadence.mjs (followup-cadence.mjs)
- FileTooLong: ./github-evidence.mjs (github-evidence.mjs)
- FileTooLong: ./invite-match.mjs (invite-match.mjs)
- FileTooLong: ./verify-portals.mjs (verify-portals.mjs)
- FileTooLong: explore/explore-provider.tsx (web/src/components/explore/explore-provider.tsx)
- Fixed-sleep synchronisation: request feedback is transition-only, quiet when closed and user-navigated (web/tests/lib/assistant-request-feedback.browser.test.mjs)
- FunctionTooLong: analyze-patterns.printSummary (analyze-patterns.mjs)
- FunctionTooLong: apply-checklist.ApplyChecklist (web/src/components/apply-checklist.tsx)
- FunctionTooLong: check-liveness.main (check-liveness.mjs)
- FunctionTooLong: claude-usage-scan.createUsageScanner (web/src/lib/claude-usage-scan.mjs)
- FunctionTooLong: employer-source.scanEmployer (web/src/lib/core/employer-source.mjs)
- FunctionTooLong: github-evidence.buildEvidence (github-evidence.mjs)
- FunctionTooLong: rank-pipeline.main (rank-pipeline.mjs)
- High: security finding (details withheld)
- High: security finding (details withheld)
- High: security finding (details withheld)
- Hotspot: assessment-log.mjs (assessment-log.mjs)
- Hotspot: check-table-freshness.mjs (check-table-freshness.mjs)
- …and 394 more
Changes since last survey
- 300 commits — 132 feature/other, 168 fixes
By area
- (root) — 160 commits
- web/src — 28 commits
- tests/providers — 9 commits
- dashboard/internal — 6 commits
- web/tests — 6 commits
- docs/SUPPORTED_JOB_BOARDS.md — 5 commits
- .agents/skills — 3 commits
- .github/workflows — 3 commits
- batch/batch-prompt.md — 3 commits
- providers/arbeitsagentur.mjs — 3 commits
- tests/helpers — 3 commits
- docs/FAQ.md — 2 commits
- lib/status-hooks.mjs — 2 commits
- modes/ar — 2 commits
- modes/interview — 2 commits
- modes/tr — 2 commits
- providers/_html-to-text.mjs — 2 commits
- providers/_http.mjs — 2 commits
- .opencode/commands — 1 commit
- batch/README.md — 1 commit
Notable commits
- fix: fix(arbeitsagentur): make angebotsart configurable so Werkstudent/Praktikum searches return results (#5788)
- fix: fix(batch): keep --company off shared ATS domains and single-dash values (#5750)
- fix: fix(batch-tailor): propagate failed worker exits (#5438)
- fix: fix(batch-tailor): run an npm-installed claude on Windows (#5919)
- fix: fix(build-cv-html): escape braces so candidate text never fills as a placeholder (#5907)
- fix: fix(build-cv-html): read a relative profile photo from the data root (#5909)
- fix: fix(build-cv-html): warn when the chosen template has no slot for a populated section (#5908)
- fix: fix(check-jd-archive): reject every localized template placeholder, not only the English one (#5926)
- fix: fix(check-liveness): --status selects the rows it names, or stops the run (#5893)
- fix: fix(cli): handle --flag=value form in discover-new-companies, company-funded, prepare-application (#5851)
- fix: fix(cli): handle --help and -h in fetch-jd and seed-fixture (#5822)
- fix: fix(close-stale): a skip gate that cannot be read is not a gate that passed (#5829)
- fix: fix(close-stale): skip rows when the interview company differs by a legal suffix (#6007)
- fix: fix(contact-extract): date a saved contact with the local day, not UTC (#5864)
- fix: fix(contacts): update the authoritative duplicate row (#5434)
- fix: fix(cv-structure): a payload that matches nothing in cv.md is unverified, not passed (#5888)
- fix: fix(cv-sync-check): match profile field values, not strings in comments (#5746)
- fix: fix(cv-visual): retry the snapshot apt fetch and fail on the real error (#6018)
- fix: fix(dashboard): a "—" or "N/A" URL cell is not a posting URL (#5872)
- fix: fix(dashboard): colour each fit-quality slice by its tier, not its position (#5874)
- …and 280 more
Architecture
- Unchanged — 0 containers · 1 contexts · 0 edges
Written by watchdog.canine.dev from the codebase's own history, inside the signed delivery this page is composed from.
Survey your own repository
career-ops-hq/career-ops was measured the same way every project in this corpus was: the same rubric, at a pinned commit, with the result published in full. Point a surveyor at a repository you know and see whether you agree with it.
About this page
- The score is its most recent published measurement, taken on 29 October 2026 at a pinned commit. It is not a live figure and does not change until the project is measured again.
- Measured at commit 91a1fd669b55d20be62349487219792a68828a19 — the exact code this score is about.
- Scored under rubric-2026.10.5 — the same rubric and the same method as every other entry in this index.
- Measured by watchdog.canine.dev using codehealth-analyzer measure-2026.10.41+preprod-baf33b0affda.