alirezarezvani/claude-skills

This system is a comprehensive, modular framework for configuring and managing specialized AI agent skills within the Claude Code environment. It provides a structured library of domain-specific capabilities spanning engineering, business strategy, compliance, and productivity, each implemented with deterministic Python tooling and strict workflow protocols. The platform enables users to orchestrate complex, multi-agent workflows, enforce quality and security standards, and maintain persistent, evidence-based memory across sessions.

Introduces a new \google-workspace-cli\ skill that provides five Python scripts (including a doctor check, auth guide, recipe runner, security audit, and output analyzer) alongside reference guides, persona profiles, and a command reference for the \gws\ CLI. The skill enables users to automate Gmail, Drive, Sheets, Calendar, Docs, Chat, and Tasks operations, with local templates for 43 recipes and 10 role-based persona bundles to streamline common administrative workflows.

engineering-team/google-workspace-cli · high confidence

The repository now includes a \.mcp.json\ file that registers the \tessl\ binary as an MCP server, enabling tools that support the Model Context Protocol to discover and invoke Tessl's MCP capabilities via a stdio transport.

(repo-wide) · high confidence

New MCP server plugins have been added for BrowserStack and TestRail, enabling AI agents to interact with these testing platforms via standard tool calls. The BrowserStack integration exposes tools to retrieve plan details, list available browsers, query builds and sessions, update session statuses, and fetch session logs. The TestRail integration provides tools to manage projects, suites, and test cases (including creating and updating cases), as well as managing test runs and submitting results. Both integrations are configured via environment variables and expose their capabilities through the Model Context Protocol SDK.

engineering-team/playwright-pro/integrations/browserstack-mcp, engineering-team/playwright-pro/integrations/testrail-mcp · high confidence

The self-improving-agent plugin now includes a new error-capture hook that monitors Bash command output for common error patterns (such as 'Error:', 'FAILED', 'ModuleNotFoundError', etc.) while excluding false positives. When an error is detected, the hook outputs a concise reminder suggesting the user save the solution using the /si:remember command or review known patterns via /si:memory-review. The hook is registered via a new hooks.json configuration file that uses the CLAUDE\_PLUGIN\_ROOT variable to ensure correct path resolution, and it is automatically installed when the plugin is added via /plugin install.

engineering-team/self-improving-agent/hooks · high confidence

The templates directory now includes an agent-template.md file that provides a standardized structure for creating new agents, including YAML frontmatter, skill integration guidelines, and workflow documentation patterns. A CLAUDE.md guide has been added to explain how to use these templates for consistent agent development across the project.

templates · high confidence

A new pre-commit hook script (karpathy-gate.sh) has been added to the engineering/karpathy-coder/hooks directory. This hook runs automatically before commits to analyze staged Python, TypeScript, and JavaScript files. It invokes a complexity checker on individual files and a diff noise analyzer on the overall change set. The tool operates in a non-blocking mode, meaning it prints warnings about code complexity and diff noise but does not prevent the commit from proceeding, serving as an awareness tool rather than a strict enforcement mechanism.

engineering/karpathy-coder/hooks · high confidence

The tech-debt-tracker skill now includes a sample codebase asset containing intentionally flawed JavaScript and Python files (frontend.js, payment\_processor.py, user\_service.py) designed to demonstrate technical debt detection capabilities, alongside a new hook log file at .claude/claudex/log that records execution events and error traps.

_.claude/claudex, engineering/skills/tech-debt-tracker/assets/sample\codebase · high confidence

The autoresearch-agent skill now provides five slash commands—/ar:setup, /ar:run, /ar:loop, /ar:ar-status, and /ar:ar-resume—enabling users to create, run, and manage autonomous optimization experiments. The skill includes scripts for setting up experiments (setup\_experiment.py), running single iterations with evaluation and git rollback (run\_experiment.py), viewing results in terminal/CSV/markdown formats (log\_results.py), and scheduling recurring loops via cron (loop skill). It supports multiple domains (engineering, marketing, content, prompts, custom) with built-in evaluators for metrics like speed, size, test pass rate, and LLM-judged content quality.

engineering/autoresearch-agent/skills · high confidence

The Executive Mentor skill now includes dedicated modules for adversarial board meeting preparation (forcing cold number mastery and anticipating hard questions), pre-mortem plan analysis (identifying assumption vulnerabilities and dependency chains), and comprehensive reference guides for board dynamics, crisis response (cash, key person, PR, legal, customer loss, fundraising), and difficult decisions (firing, layoffs, pivots). It also introduces Python-based decision analysis tools (weighted scoring with sensitivity testing) and stakeholder mapping to support rigorous executive decision-making.

c-level-advisor/executive-mentor/skills · high confidence

The \.vibe/skills/claude-skills\ directory has been significantly expanded by adding hundreds of new symbolic links that mirror skills from various source domains (such as \agent-launcher\, \business-growth\, \c-level-advisor\, \engineering-team\, and \compliance-os\). This change ensures that the Claude-specific skills index includes a comprehensive set of capabilities from these areas, making them discoverable and usable within the Claude environment.

.vibe · high confidence

The .codex/skills directory now includes a significantly larger set of symbolic links pointing to skill definitions across various domains (engineering, marketing, compliance, etc.). This update adds support for numerous new skills such as a11y-audit, chaos-engineering, feature-flags-architect, google-workspace-cli, and kubernetes-operator, making them available for use within the Codex environment.

.codex/skills · high confidence

The .hermes directory now exposes a comprehensive set of agent-launcher, business-growth, business-operations, c-level-advisor, commercial, compliance-os, and engineering-team skills as symlinks within the claude-skills structure. This change registers the agent-launcher domain and refreshes mirror trees, allowing users to access these specialized agent capabilities directly through the Claude integration layer.

.hermes · high confidence

Adds the AgentHub plugin, a new engineering skill that enables spawning N parallel AI agents to compete on the same task using isolated git worktrees. The skill includes a coordinator protocol (init, dispatch, monitor, evaluate, merge), a message board for inter-agent communication, and predefined agent templates (optimizer, refactorer, test-writer, bug-fixer). It is supported by Python scripts for session management, DAG analysis, result ranking, and board operations, along with documentation on coordination strategies and git DAG patterns.

engineering/agenthub/skills · high confidence

Compliance OS is now available as a meta-orchestrator for multi-framework compliance programs, enabling teams to configure applicable frameworks, compute cross-framework control overlap, simulate internal audits, and consolidate evidence into a unified pool. The update adds a plugin manifest and README, along with new persona agents for the Compliance Officer (orchestrator), ISO 27001 ISMS, ISO 13485 QMS, GDPR DPO, ISO 42001 AIMS, and EU AI Act compliance. These agents route deep-dive work to existing specialist skills while providing structured workflows, output standards, and integration points for cross-framework reuse and audit readiness.

compliance-os · high confidence

Adds the \llm-wiki\ plugin, which turns an LLM CLI (Claude Code, Codex, Cursor, etc.) into a disciplined wiki maintainer for Obsidian. Instead of stateless RAG, the plugin incrementally ingests sources into a persistent, interlinked markdown vault, maintaining entity/concept pages, cross-references, and a living synthesis. It ships with five slash commands (\/wiki-init\, \/wiki-ingest\, \/wiki-query\, \/wiki-lint\, \/wiki-log\), three sub-agents (ingestor, librarian, linter), and eight Python standard-library tools to handle vault bootstrapping, source ingestion, index regeneration, health checks, and BM25 search.

engineering/llm-wiki · high confidence

The \research/pulse\ plugin (v2.9.0) adds a new capability to take the pulse of any topic across Reddit, Hacker News, the open web, and optionally X/Twitter within a configurable recent window (default 30 days). It enforces a structured intake process to clarify topic specificity, angle, time window, and platform scope before searching. The skill runs Phases 1–3 (Reddit, HN, Web) in parallel with a 1 q/sec rate limit per platform, and includes optional Phase 4 for X/Twitter data via local export, Grok, API, or browser automation. All outputs are synthesized into a single briefing with citations, engagement metrics, and cross-platform pattern analysis (consensus, controversy, pain points, excitement, emerging trends, gaps). The implementation includes strict source discipline: only sources returned by the session's tool calls are cited, training knowledge is excluded from primary findings, and a three-count audit log (queries sent, sources received, sources cited) is surfaced in the synthesis. The skill also includes robust failure handling: retry-once-after-3s, and stop-after-3-consecutive-failures across all sources.

research/pulse · high confidence

Adds the \skillopt-sleep\ plugin, a vendored and hardened copy of Microsoft's SkillOpt engine, enabling the Claude agent to harvest past sessions, mine recurring tasks, and replay them offline to consolidate validated memory and skills. The plugin includes a CLI (\/skillopt-sleep\) with actions like \run\, \dry-run\, \adopt\, and \schedule\, along with shell launchers and a session-end hook. It features safety hardening including secret redaction across all outputs, shell-quoting for crontab entries, and restrictive file permissions, while removing hardcoded internal Azure infrastructure and dead configuration keys.

engineering/skillopt-sleep · high confidence

The engineering/agent-harness area now provides a complete system for turning any domain folder of skills into a bounded agentic loop. This includes a plugin manifest, a \/cs:harness\ command, and a \harness-runner\ agent that drive a state machine (compile goal → execute → verify → close). The system enforces machine-run verification, retry caps, and human escalation on exhausted budgets, refusing to close until all tasks are verified. It ships with 18 committed per-domain manifests (e.g., business-growth, c-level-advisor) and a JSON schema, ensuring that agents pick up goals and drive them to a verified close across specific domains.

engineering/agent-harness · high confidence

This change adds a new \agent-memory\ plugin that replaces the flat, always-injected \CLAUDE.md\ memory model with a four-tier ladder (L0 raw transcripts, L1 candidate atoms, L2 project context, L3 stable persona). Memory is promoted based on recurrence across distinct sessions and days rather than asserted confidence, and no fact reaches a committed file without explicit human adoption via the \/cs:memory adopt\ command. The system uses three hooks (\SessionStart\, \UserPromptSubmit\, \SessionEnd\) to capture, recall, and stage memory, with strict gates that refuse promotion for redacted or contested claims.

engineering/agent-memory · high confidence

Adds a new chaos-engineering skill that provides an end-to-end discipline for designing, running, and learning from chaos experiments. It includes three Python tools (experiment\_designer, blast\_radius\_calculator, experiment\_postmortem), reference documentation on chaos principles and attack taxonomies, and templates for experiment plans and postmortems. The skill enforces safety by requiring hypotheses, steady-state metrics, blast radius calculations, and abort criteria, and integrates with other skills like feature-flags-architect and kubernetes-operator.

engineering/chaos-engineering · high confidence

Adds a new engineering skill that serves as a date-stamped study companion for the 2016 Deep Learning textbook. The skill provides a navigable knowledge base with 20 chapter files, a glossary, and a delta layer that updates 2016-era advice against 2026 practice (covering topics like double descent, AdamW, and transformers). It includes four standard-library-only CLI tools for planning reading paths, diagnosing training failures, planning model capacity, and calculating parameter/FLOP counts, along with agent and slash-command integrations to help users study, teach, or verify the currency of the book's recommendations without reproducing its copyrighted text.

engineering/deep-learning-book · high confidence

Adds the \human-gate\ skill, a stdlib-Python tool that creates a single-file, offline HTML review page for Markdown or HTML artifacts and collects feedback as a structured \batch.v1\ JSON artifact rather than unstructured chat prose. The system enforces a closing gate with seven integrity rules (G1–G7), refusing to mark work as done if no review exists, blocking items remain open, the reviewer is unnamed, the sidecar changed after collection, the round cap is exhausted, a waiver lacks a reason, or integrity problems like mistyped severities are present. It supports a non-blocking loop with a configurable round cap (default 5) that escalates on exhaustion, and includes a \cs-human-gate\ agent command to orchestrate the open/status/collect/close workflow.

engineering/human-gate · high confidence

The markdown-html area now provides a complete domain for converting long markdown files into single-file, lightly-interactive HTML. It includes a markdown-html-orchestrator that classifies input and routes to one of three converter skills: md-document for long-form content (sticky TOC, search, code-copy, scrollspy), md-review for code reviews (2-column diff with severity-tagged annotations), and md-slides for presentations (keyboard navigation, presenter mode, print-to-PDF). All converters share a design-system skill that runs a one-time onboarding wizard to capture brand colors, typography, and layout preferences, enforcing WCAG AA contrast and storing the configuration for consistent rendering. The domain refuses conversions for short markdown (\<100 lines), missing onboarding, or unwritable output paths, and outputs are strictly single-file HTML with only Google Fonts and optional Prism.js as external dependencies.

markdown-html · high confidence

Adds the \memory-engineering\ plugin (v2.11.2), which provides a four-lens workflow to price, choose, audit, and gate agent memory systems. It includes four deterministic stdlib scripts: \memory\_cost\_profiler.py\ to split construction vs. query spend and report cost per correct answer; \memory\_architecture\_picker.py\ to score long-context, flat RAG, structure-augmented RAG, and agentic paradigms against constraints; \memory\_density\_auditor.py\ to classify records as FACT/SKILL/LOG/PROSE and flag duplicates or staleness; and \forgetting\_policy\_linter.py\ to enforce explicit forgetting rules (F1) and prevent auto-merging of contradictions (F4). The skill is accessible via the \cs-memory-engineer\ agent and the \/cs:memory-engineering\ and \/cs:forgetting-audit\ commands, and ships with a seven-question forcing worksheet, design spec templates, and reference materials citing Stanford, Microsoft, Anthropic, and Nvidia research.

engineering/memory-engineering · high confidence

The repository now includes a new \standards/\ directory that centralizes development guidelines for agents and skills. This library defines specific protocols for communication (direct, actionable responses), code quality (PEP 8, type hints, zero-defect handoff), documentation (living docs, markdown structure), git workflows (conventional commits, semantic versioning), and security (secret detection, input validation). These standards serve as the reference implementation for how new agents and skills should be structured, tested, and committed.

standards · high confidence

The research skill now uses a deterministic keyword-matching classifier to route queries to specialist domains (pulse, grants, litreview, syllabus, patent, dossier, deepread) or execute a structured fallback workflow when no specialist matches. The orchestrator requires explicit user confirmation for routing decisions, supports output in markdown or DOCX, and logs all routing actions, overrides, and search metrics for auditability.

research/research/skills · high confidence

The Karpathy Coder capability has been consolidated from multiple scattered plugins into a single, self-contained skill located at \engineering/karpathy-coder/skills/karpathy-coder/\. This update introduces four Python-based analysis scripts—\assumption\_linter.py\, \complexity\_checker.py\, \diff\_surgeon.py\, and \goal\_verifier.py\—that enforce Andrej Karpathy’s four coding principles (Think Before Coding, Simplicity First, Surgical Changes, and Goal-Driven Execution). Users can now run these tools directly via the \/karpathy-check\ slash command or integrate them into pre-commit hooks and CI pipelines to automatically detect over-engineering, diff noise, hidden assumptions, and weak verification steps in their code.

engineering/karpathy-coder/skills · high confidence

The productivity plugin now includes a new 'Capture' skill that transforms unstructured brain dumps into a structured, four-section actionable system (Projects & Ideas, Tasks, Connections, How I Can Help) with zero information loss. The skill automatically triggers on explicit phrases like 'brain dump' or implicit unstructured text pastes, preserving the user's original voice without corporate-ifying the language. It uses three Python helper scripts (dump\_classifier, complexity\_estimator, workspace\_inventory) to classify items, determine output format (full 4-section vs. compressed for small dumps), and verify real workspace connections via Glob/Grep, ensuring no fabricated links. The skill asks at most one clarifying question if an item is ambiguous between a task and a project, and waits for user approval before executing any offers.

productivity/email · high confidence

Adds a new \productivity/roast\ skill that pressure-tests business ideas before building by convening a five-angle adversarial panel (Critic, Champion, Analyst, Investigator, Customer) in parallel. The skill uses three deterministic Python tools (\brief\_builder.py\, \verdict\_synthesizer.py\, \cheapest\_test\_designer.py\) to synthesize a weighted GO / RESHAPE / KILL verdict with explicit confidence and a falsifiable 48-hour test, accessible via the \/cs:roast\ command.

productivity/roast · high confidence

Added the \grill-with-docs\ skill, which interrogates design plans against the project's existing domain language (\CONTEXT.md\) and recorded architectural decisions (\docs/adr/\). This tool sharpens terminology and records decisions inline, differentiating itself from the plan-only \grill-me\ skill by grounding its analysis in the codebase's documented context. The skill includes a \cs-grill-with-docs\ persona agent, a \/cs:grill-with-docs\ slash command, and three Python validation tools (a \CONTEXT.md\ linter, an ADR scanner, and a glossary-to-code consistency checker) to ensure the glossary and decision records remain accurate and up-to-date during the grilling process.

engineering/grill-with-docs · high confidence

Adds a new \meetings\ plugin (v2.11.2) that enforces a three-stage discipline for individual meeting management: a cost gate that prices meetings in real dollars and requires a decision, agenda, and owner (verdicts: ASYNC, NOT-READY, or MEET); a timeboxed agenda builder that prioritizes decision topics and enforces a closing actions-recap buffer; and a deterministic action-item extractor that converts raw notes into an owner-grouped checklist, flagging any items missing an owner (ORPHAN) or a due date (NO-DUE). The plugin includes CLI scripts (\meeting\_cost\_calculator.py\, \agenda\_builder.py\, \action\_item\_extractor.py\), agent instructions, and reference materials based on meeting-science canon.

productivity/meetings · high confidence

Added the 'reflect' skill, a pure-reasoning productivity tool that pauses execution to reassess conversation direction, assumptions, and bias. It operates as a sibling to the 'capture' skill, using a 5-dimension analysis framework (Macro Perspective, Gap Analysis, Reflective Inquiry, Bias Check, Contextual Alignment) to detect drift and cognitive biases like confirmation or sunk cost. The skill triggers on explicit phrases (e.g., 'step back', 'zoom out') or implicit signals (10+ turns of detail without strategic check-in), runs heuristic Python scripts for bias detection, and outputs a flowing-prose recommendation to Continue, Pivot, or Pause, ensuring honest validation without manufactured problems.

productivity/reflect · high confidence

This location introduces the \write-a-skill\ skill, derived from Matt Pocock's MIT-licensed work, which enables users to author new agent skills using a structured 3-phase workflow (Gather, Draft, Review). To enforce quality standards, the skill includes three Python validation tools (\skill\_description\_validator.py\, \skill\_structure\_validator.py\, \skill\_review\_checklist\_runner.py\) that programmatically check for description length, third-person voice, trigger phrases, and file structure. It also adds a \cs-skill-author\ persona agent and a \/cs:write-a-skill\ slash command to provide a forcing-question interrogation gate before any new skill is committed.

engineering/write-a-skill · high confidence

A new 'Agent Designer' skill has been added to the engineering toolkit, providing a structured workflow for designing, schema-generating, and evaluating multi-agent systems. It includes three core components: an Agent Planner for selecting orchestration patterns (such as Supervisor, Swarm, or Pipeline) and generating implementation roadmaps; a Tool Schema Generator that produces validated, provider-ready schemas for both OpenAI and Anthropic formats; and an Agent Evaluator for analyzing execution logs to identify performance bottlenecks, cost inefficiencies, and critical errors. The skill also ships with comprehensive reference documentation on architecture patterns, tool design best practices, and evaluation methodologies, along with sample assets for quick start.

engineering/skills · high confidence

The Agile Product Owner skill now includes a comprehensive documentation suite and an automated user story generator. Users can leverage structured templates for sprint planning and user stories, along with reference guides for capacity calculation, backlog prioritization (including WSJF), and INVEST criteria validation. Additionally, a new Python script automates the generation of INVEST-compliant user stories and acceptance criteria from epic definitions, streamlining the creation of backlog items.

product-team/agile-product-owner/skills · high confidence

Adds a new persona skill that pressure-tests ventures and ideas using Marc Andreessen's frameworks, delivering verdicts such as BUILD-POUR-FUEL, MARKET-FIRST-DERISK, or KILL-OR-REPICK-MARKET based on a market-first evaluation. It includes a daily productivity routine (3x5 card and Anti-Todo list) and ships with three Python scripts to score market viability, product/market fit signals, and manage the daily card, alongside reference documents and worked examples.

productivity/andreessen/skills · high confidence

Added two new skills to the project-management toolkit: 'atlassian-admin' and 'atlassian-templates'. The atlassian-admin skill provides expert guidance for managing Jira, Confluence, Bitbucket, and Trello, including user provisioning, permission scheme design, SSO configuration, and security hardening, supported by a Python-based permission audit tool and reference checklists. The atlassian-templates skill enables the creation and management of reusable Jira and Confluence templates, featuring a scaffolder script to generate Confluence storage-format markup and a governance framework for template lifecycle management.

project-management/skills · high confidence

Added new agent definitions for the cs-ceo-advisor and cs-cto-advisor in the agents/c-level directory. The CEO advisor provides strategic guidance on leadership, board management, and investor relations, while the CTO advisor focuses on technology strategy, engineering team scaling, and architecture decisions. Both agents are configured to use the opus model and integrate with specific Python analysis tools and knowledge bases located in the c-level-advisor skills directory.

agents/c-level · high confidence

The scripts directory now includes a suite of new validation tools and installation utilities. New linters enforce content quality: check\_frontmatter.py validates YAML frontmatter (Gate G10), check\_model\_freshness.py flags retired model identifiers (Gate G7), check\_paths.py detects unresolvable path references (Gate G1), check\_skill\_names.py prevents shadowing of built-in commands, check\_plugin\_json.py ensures plugin manifests match the Claude Code spec, and check\_dual\_publish.py verifies that dual-published skill pairs remain in sync. Additionally, audit\_skills.py provides a repo-wide skill audit runner, while codex-install.sh and codex-install.bat offer cross-platform installation of skills to the local Codex directory.

scripts · high confidence

A new standalone plugin, \chief-ai-officer-advisor\, has been added to provide strategic advisory for C-level executives. It covers four key decision areas: model build-vs-buy analysis (API vs. fine-tune vs. in-house), AI risk classification under the EU AI Act and US state regulations, API-to-self-hosted cost economics, and AI team organizational evolution. The plugin includes Python scripts for automated calculations (\model\_buildvsbuy\_calculator.py\, \ai\_risk\_classifier.py\, \ai\_cost\_economics.py\) and detailed reference documentation. It is dual-published, meaning it is also bundled within the \c-level-skills\ plugin, with content kept in sync via a dedicated script.

c-level-advisor/chief-ai-officer-advisor · high confidence

Adds a standalone plugin for Chief Customer Officer advisory, providing strategic guidance on retention decomposition (GRR vs NRR), customer segmentation (4-tier framework with ICP fit scoring), CS team coverage models (pooled vs named CSMs), and CS org evolution. The skill includes four Python scripts for automated analysis—retention\_decomposition\_analyzer.py, customer\_segmentation\_designer.py, cs\_coverage\_calculator.py, and a fourth implied by the SKILL.md—and detailed reference documentation for each decision area, enabling founders and CCOs to make data-driven decisions on customer success strategy without duplicating tactical CS management skills.

c-level-advisor/chief-customer-officer-advisor · high confidence

This update introduces two new capabilities in the C-level advisor skills. First, the Company Architect (arquiteto-de-empresa) skill guides founders through a 12-phase interview to build a company as code, outputting a conformant Open Knowledge Format (OKF) bundle of cross-linked Markdown files; it includes three standard-library Python tools (scaffold\_bundle, okf\_linter, index\_generator) to automate bundle creation, validation, and index regeneration. Second, the Inter-Agent Protocol defines how C-suite agents communicate, specifying an invocation syntax (\[INVOKE:role\|...\]), a strict response format, and hard rules for loop prevention (no self-invocation, max depth of 2, no circular calls) and isolation during board meeting phases, along with a broadcast pattern for crisis scenarios and a canonical decision-memory layout.

c-level-advisor/skills · high confidence

Adds a new \docker-development\ skill that provides slash commands (\/docker:optimize\, \/docker:compose\, \/docker:security\) for Dockerfile optimization, docker-compose orchestration, and container security auditing. The skill includes a Python-based static analyzer (\dockerfile\_analyzer.py\) to detect anti-patterns and a compose validator (\compose\_validator.py\) to check for best practices, along with reference guides for multi-stage builds, network isolation, and security hardening.

engineering/docker-development · high confidence

Adds a new standalone plugin for Regulation (EU) 2024/1689 (the EU AI Act) designed for compliance teams. The plugin provides three deterministic Python tools to classify AI systems by risk tier (prohibited, high-risk, limited-risk, minimal-risk), plan conformity assessments (Module A vs. Module H) for high-risk systems, and track organizational obligations (provider, deployer, importer, distributor). It includes comprehensive reference documentation mapping AI Act requirements to ISO 42001, NIST AI RMF, and GDPR, and covers specific obligations for General-Purpose AI (GPAI) models under Articles 51–55.

ra-qm-team/compliance-team-eu-ai-act · high confidence

The engineering/feature-flags-architect skill is now available, providing a structured lifecycle for feature flags (classify, ship, ramp, retire) to prevent flag debt. It includes three Python tools: a flag debt scanner to identify stale flags, a rollout planner to generate phased schedules (ring, linear, log, cohort), and a kill-switch auditor to verify documentation. The skill also ships with reference documentation on flag taxonomy, provider comparisons (LaunchDarkly, GrowthBook, Statsig, Unleash, Flipt, DIY), and rollout strategies, along with a flag request template to standardize new flag creation.

engineering/feature-flags-architect · high confidence

Adds a new engineering skill that enables AI agents to scaffold, review, and audit production-grade Helm charts. The skill provides slash commands (/helm:create, /helm:review, /helm:security) to automate chart structure generation, analyze templates for anti-patterns and missing labels, and enforce security best practices like non-root containers and network policies. It includes reference documentation for standard chart layouts and values design, along with Python scripts for static analysis of chart directories and validation of values.yaml files.

engineering/helm-chart-builder · high confidence

The engineering/hivemind directory now contains a complete orchestration skill that uses Claude Code as the brain and the opencode CLI as a swarm of free, disposable workers. It introduces slash commands (/hive, /oc, /swarm, /review-panel, /research-sweep, /migration, /test-fleet) and agent definitions (scout, coder, tester) to delegate tasks in parallel with worktree isolation, plus scripts (oc-worker, oc-status, oc-aggregate) to spawn workers, track progress, and aggregate findings. A benchmark runner (run-bench) compares Claude solo, opencode solo, and orchestrated swarm configurations, while a .gitignore excludes runtime state and bench results. This location provides the skill assets, agent prompts, slash commands, and orchestration scripts; other areas handle the broader platform integration.

engineering/hivemind · high confidence

Adds a standalone plugin providing three deterministic tools for internal AI Management System (AIMS) audits: an AIMS gap analyzer for Clauses 4–10 coverage scoring, an AI risk register builder mapping risks to Annex A controls per ISO 23894, and an internal audit scheduler for 12-month planning. The package includes Python scripts, a detailed skill definition, and reference documentation covering control catalogs, implementation guides, and cross-framework mappings to EU AI Act and NIST AI RMF.

ra-qm-team/compliance-team-iso42001 · high confidence

A new \kubernetes-operator\ skill is available to help teams build and audit Kubernetes Operators correctly. It ships three Python tools—\crd\_validator.py\ for CRD design checks, \reconcile\_lint.py\ for Go controller anti-pattern detection, and \operator\_capability\_audit.py\ for scoring against OperatorHub Capability Levels—along with a \/operator-audit\ slash command, production-ready CRD and Go controller templates, and reference documentation on the operator pattern, CRD design, and reconcile loops.

engineering/kubernetes-operator · high confidence

Added a new \grants\ skill in the research pack that guides clinical researchers through a structured intake to produce a strategic NIH funding overview as an editable .docx. The skill performs a 5-facet Consensus positioning analysis, maps the research to NIH institutes and study sections via RePORTER POST queries, discovers relevant NOSIs, and generates scope-aware mechanism recommendations (e.g., R01, K-award, F-series) based on career stage, project scope, and preliminary data. It enforces strict source discipline and audit logging, and includes a mandatory program officer recommendation. The skill is NIH-only; non-NIH funders are explicitly out of scope.

research/grants · high confidence

Adds a new \notebooklm\ skill in the \research/\ directory that automates Google NotebookLM via browser automation. The skill supports four core actions: reading/extracting from existing notebooks, adding sources (URLs, text, files, Google Docs), generating Studio outputs (Audio/Video Overviews, Mind Maps, Reports, Slides, etc.), and creating new notebooks. It enforces a screenshot-first, semantic-find discipline, requires mandatory custom prompts for Studio outputs to avoid mediocre defaults, and uses a fire-and-notify pattern for slow asynchronous generations (like Audio Overviews) to prevent session timeouts. The skill fails gracefully if browser automation is unavailable or if a login wall is detected.

research/notebooklm · high confidence

The \c-level-advisor/arquiteto-de-empresa/skills/arquiteto-de-empresa/scripts\ directory now includes three new Python utilities to support the Open Knowledge Format (OKF) bundle structure. \scaffold\_bundle.py\ generates the initial folder skeleton (including phase dashboards and logs) for a company, with optional product and tech modules. \index\_generator.py\ automatically regenerates concept tables within \index.md\ files based on sibling markdown files, supporting both dry-run previews and direct writes. \okf\_linter.py\ validates bundle conformance by checking for required frontmatter types, controlled vocabulary usage, and valid relative links, reporting errors and warnings.

c-level-advisor/arquiteto-de-empresa/skills/arquiteto-de-empresa/scripts · high confidence

The \research-ops\ area introduces an enterprise Research Operations domain (v2.9.0) comprising four sub-skills—\clinical-research\ (study design, endpoint selection, sample-size/power, phase-gate feasibility), \research-finance\ (R&D program budgeting, burn/runway, capitalize-vs-expense routing), \market-research\ (TAM/SAM/SOM sizing, survey sampling, segmentation), and \product-research\ (study design, saturation, insight synthesis)—orchestrated by \research-ops-skills\. Each sub-skill includes per-skill onboarding (\onboard.py\) and a customization loader (\config\_loader.py\) that pre-configures tools via project/global/defaults precedence, plus an isolated, opt-in autoresearch bridge (\ar\_evaluator.py\) that connects to \engineering/autoresearch-agent\ for iterative optimization without cross-skill coupling. The orchestrator routes inquiries to the appropriate lane using deterministic signal classification and forks context to keep heavy intake separate, while enforcing hard rules that outputs are estimates with named human owners, assumptions are always surfaced, and single-source claims are flagged as anecdotes rather than insights.

research-ops · high confidence

Introduces the SLO Architect skill, which enforces Google SRE Workbook discipline for Service Level Objectives. The skill provides three Python tools: an SLO designer that generates structured definitions with required fields, an error-budget calculator that computes multi-window burn-rate alert thresholds (fast, slow, and ticket) with PromQL-shaped output, and an SLO reviewer that audits existing definitions for common anti-patterns like targets that are too high, windows that are too short, or CPU usage used as an SLI. It also includes asset templates for SLO YAML and error budget policies, along with reference documentation on SLI design and composition with other skills like feature-flags-architect and chaos-engineering.

engineering/slo-architect · high confidence

Adds the snowflake-development plugin, providing a comprehensive skill for Snowflake SQL, data pipelines (Dynamic Tables, Streams+Tasks), Cortex AI functions, Snowpark Python, and dbt integration. The update includes a new plugin.json manifest (version 2.9.0) and a Python helper script (snowflake\_query\_helper.py) that generates MERGE upserts, Dynamic Table DDL, and RBAC grant statements. It also introduces detailed reference documentation covering SQL best practices, Cortex AI/Agent specifications, and troubleshooting guides to help users avoid common pitfalls like deprecated function names and incorrect TO\_FILE usage.

engineering-team/snowflake-development · high confidence

Adds the statistical-analyst plugin (version 2.9.0) which provides three Python tools for statistical analysis: hypothesis\_tester.py for running Z-tests, t-tests, and Chi-square tests; sample\_size\_calculator.py for pre-launch experiment sizing; and confidence\_interval.py for computing Wilson score and t-based confidence intervals. The plugin includes a SKILL.md guide for integrating these tools into Claude workflows and a reference document detailing the underlying frequentist statistical concepts.

engineering/statistical-analyst · high confidence

Adds a new 'terraform-patterns' skill that provides agent capabilities for designing Terraform modules, managing state backends, auditing security configurations, and enforcing infrastructure-as-code best practices. The skill includes slash commands (/terraform:review, /terraform:module, /terraform:security) and bundled Python utilities (tf\_module\_analyzer.py, tf\_security\_scanner.py) to analyze code structure, detect anti-patterns, and scan for security vulnerabilities like hardcoded secrets or overly permissive IAM policies.

engineering/terraform-patterns · high confidence

This change introduces the \vpe-advisor\ plugin, providing strategic advisory capabilities for VP of Engineering roles. The skill covers four key areas: delivery throughput analysis using DORA metrics to identify bottlenecks, engineering hiring funnel calculation to diagnose leakage and pipeline gaps, engineering team structure design based on squad/tribe models and manager-trigger thresholds, and production discipline auditing for on-call, incident response, and SLOs. It includes Python scripts (\delivery\_throughput\_analyzer.py\, \eng\_hiring\_funnel\_calculator.py\, \eng\_team\_structure\_designer.py\) for automated analysis and detailed reference documentation. The plugin is version 2.9.0 and is designed to be standalone or bundled within the \c-level-skills\ suite.

c-level-advisor/vpe-advisor · high confidence

A new 'a11y-audit' skill has been added to the skills directory, providing a complete accessibility audit pipeline for scanning, fixing, and verifying WCAG 2.2 Level A and AA compliance across React, Next.js, Vue, Angular, Svelte, and plain HTML codebases. This skill includes a three-phase workflow (Scan, Fix, Verify) with framework-specific fix code generation, severity classification (Critical, Major, Minor), and stakeholder-ready compliance reporting. It ships with Python-based tooling (\a11y\_scanner.py\ and \contrast\_checker.py\) for automated scanning and color contrast validation, along with comprehensive reference documentation covering ARIA patterns, framework-specific accessibility fixes, CI/CD integration examples (GitHub Actions, GitLab CI, Azure DevOps), and WCAG 2.2 criteria references. Sample components and expected outputs are included to demonstrate the skill's capabilities.

engineering-team/a11y-audit/skills · high confidence

A new \agent-decision-receipts\ skill has been added to the skills directory, enabling the minting and offline verification of tamper-evident receipts for consequential agent actions (such as deploy, delete, or pay). The skill delegates cryptographic signing to the \openagentontology\ package, supporting Ed25519 and post-quantum signatures (ML-DSA-65 and SLH-DSA) to ensure long-term evidence integrity. It includes a \build\_action\_manifest.py\ script for creating validated, ASCII-safe action manifests and provides a structured workflow for deciding when to receipt, minting the receipt, and verifying it from the certificate alone without network or database access. This capability supports compliance with record-keeping requirements like EU AI Act Article 12.

ra-qm-team/skills · high confidence

The agent-launcher area now ships a new Claude Code plugin (version 2.11.2) that scaffolds and launches Claude Managed Agents (CMA) in the user's own Anthropic account. Every session starts with a goal (./my-agent/goal.json), which is compiled into a single-pass workflow, a bounded grade→iterate loop, or a recurring POSIX-cron deployment loop. The plugin provides six skills (orchestrator, interview, stage-launch, grade-iterate, run-without-you, wrap-up), four agents, eight /cs:\* commands, and 18 stdlib-only deterministic scaffolder tools that never make network or API calls; live launches are emitted as BYOK curl scripts. An opt-in SessionStart hook (gated by AGENT\_LAUNCHER\_SESSION=1) surfaces the current goal so multi-session launches resume cleanly, and validators enforce CMA limits (e.g., max 20 skills per session, max 20 iterations).

agent-launcher · high confidence

Introduces the book-to-skill plugin, which converts books, documentation folders, and source collections (PDF, EPUB, DOCX, HTML, Markdown, RST, AsciiDoc, RTF, MOBI/AZW) into structured agent skills. The tool produces a resident core (SKILL.md) with core frameworks and indexes, on-demand chapter files, a glossary, a patterns file, and a decision cheatsheet. It includes four command-line tools for extraction, validation, token budget estimation, and plugin emission, along with an agent persona and CLI commands to drive the conversion workflow. The implementation is derived from an upstream library with specific deviations for safety, rights gating, and Claude Code integration.

engineering/book-to-skill · high confidence

The \business-growth/skills\ area now includes a router and three specialized skill plugins: \customer-success-manager\ (with Python tools for health scoring, churn risk, and expansion opportunities, plus QBR and onboarding templates), \contract-and-proposal-writer\ (for jurisdiction-aware contracts, NDAs, and SOWs), and \revenue-operations\ (for pipeline and forecast analysis). These skills provide structured workflows, reference guides, and sample data to support customer retention, legal document generation, and revenue planning.

business-growth/skills · high confidence

Added new Python-based skills under the business-operations and internal-comms domains to support capacity planning, hiring sequencing, utilization analysis, and structured internal change management. The capacity-planner skill provides Erlang-C queueing math for sizing ops teams against demand and SLA targets, a 12-month hiring sequencer that accounts for ramp time and attrition, and a utilization analyzer that flags overloaded or unbalanced teams. The internal-comms skill suite generates Kotter 8-step compliant change announcements with tone validation, builds a 7-touchpoint Prosci-aligned communication calendar, and fills a 4-artifact comms package (pre-comm, announcement, FAQ, follow-up) mapped to ADKAR stages.

python · high confidence

Three new engineering skills—caveman, grill-me, and handoff—are now available, each derived from Matt Pocock’s MIT-licensed skills and enhanced with stdlib Python tooling, deep reference docs, and cs-\* persona agents with slash commands. Caveman provides ultra-compressed communication (dropping filler, articles, and pleasantries while preserving technical accuracy) with a compressor, token-savings estimator, and linter to enforce rules and quantify savings. Grill-me acts as a relentless plan-and-design interrogator that walks decision trees one branch at a time, using a decision-tree extractor, question generator, and session-state tracker to maintain focus and track progress. Handoff (location specified but content not shown in this excerpt) follows the same pattern of derivation and enhancement. All skills preserve the original authors’ voice and rules verbatim per MIT, and include attribution metadata in authoring-notes.json and plugin.json.

engineering/caveman, engineering/grill-me, engineering/handoff · high confidence

A new \channel-economics\ skill has been added to the commercial skills library to help Head of Commercial, RevOps, and VP Sales teams evaluate the profitability of direct versus partner-led sales channels. This skill provides a structured workflow to compute fully-loaded cost-to-serve, analyze ROI across cash, LTV, and marginal lenses, and optimize channel mix under strategic constraints. It includes three Python scripts (\cost\_to\_serve\_calculator.py\, \channel\_roi\_analyzer.py\, and \channel\_mix\_optimizer.py\) that generate deterministic verdicts (DOUBLE-DOWN, MAINTAIN, DEFUND, EXIT) and sensitivity-tested recommendations, along with data templates and reference materials on common anti-patterns like inconsistent overhead allocation and influence-vs-source attribution errors.

commercial/skills · high confidence

A new 'dossier' skill has been added to the research pack, enabling decision-grade entity research on companies, people, nonprofits, or government organizations. Unlike generic profiles, this skill forces users to state a specific hypothesis upfront and allocates at least 30% of the search budget to disconfirming evidence, ensuring the output tests rather than confirms prior beliefs. The skill produces an editable Word document (.docx) containing a verdict on the hypothesis, a 12-month activity timeline, network and reputation signals, red flags, and specific conversation hooks tied to findings. It utilizes WebSearch, WebFetch, and free APIs (SEC EDGAR, GitHub, ProPublica) as core workhorses, with optional BYOK MCPs for enhanced coverage. Supporting scripts handle citation tracking, source tier classification, and disconfirming evidence balance enforcement to maintain research integrity.

research/dossier/skills · high confidence

A new \deep-work\ skill has been added to the productivity suite, providing a structured approach to planning a focused day. It includes three Python scripts: \shallow\_work\_auditor.py\ to classify tasks and enforce a shallow-work budget, \time\_block\_planner.py\ to generate an energy-first schedule with deep blocks, buffers, and a 4-hour deep-work cap, and \focus\_session\_logger.py\ to track weekly deep-work hours and streaks. The skill is supported by reference materials on deep vs. shallow work, time-blocking mechanics, and a shutdown ritual checklist to close out the day.

productivity/deep-work/skills · high confidence

Three new skills have been added to the engineering toolkit. The Adversarial Reviewer skill introduces a structured code review process using three hostile personas (Saboteur, New Hire, Security Auditor) to force perspective shifts and eliminate rubber-stamp approvals, outputting BLOCK/CONCERNS/CLEAN verdicts. The AI Security skill provides methodology and a Python threat scanner tool for assessing LLM systems against prompt injection, jailbreaks, model inversion, and data poisoning, mapping findings to MITRE ATLAS techniques. The AWS Solution Architect skill guides users through designing serverless and multi-tier architectures, generating Infrastructure-as-Code templates (CloudFormation, CDK, Terraform), and analyzing cost optimization opportunities.

engineering-team/skills · high confidence

The autoresearch-agent now includes a suite of new evaluators in the \evaluators\ directory to measure specific aspects of code and content. These include \benchmark\_size\, \benchmark\_speed\, and \build\_speed\ for performance metrics; \llm\_judge\_content\, \llm\_judge\_copy\, and \llm\_judge\_prompt\ for evaluating content, marketing copy, and prompt quality using external LLM CLI tools; \memory\_usage\ for tracking peak memory consumption; and \test\_pass\_rate\ for measuring test suite success rates.

engineering/autoresearch-agent/evaluators · high confidence

Added two new expert skills to the platform. The Data Quality Auditor skill provides a systematic workflow for profiling datasets, detecting anomalies, and generating remediation plans, supported by Python scripts for profiling, missing value analysis, and outlier detection. The Apple HIG Expert skill enables auditing and designing iOS, macOS, visionOS, and watchOS interfaces against the Human Interface Guidelines (including the Liquid Glass design language), featuring a compliance checker script for contrast and tap-target validation, along with detailed reference guides for accessibility, visual design, and platform-specific ergonomics.

(repo-wide) · high confidence

A new productivity skill, fable-goal, has been added to convert user rambling into a polished, copy-paste /goal prompt designed for autonomous sessions. The skill (SKILL.md) guides the model to extract intent, verify resources, and structure the output with specific components like a verification loop and autonomy directive. It includes a supporting rationale document (references/goal\_prompt\_patterns.md) explaining the design principles and failure modes. Additionally, a Python self-check script (scripts/goal\_prompt\_self\_check.py) is provided to mechanically verify that generated prompts meet structural requirements such as word count, presence of a goal line, and verification language.

productivity/fable-goal/skills · high confidence

Added a new finance skills bundle under finance/skills/ containing a router, a financial-analyst skill, and a saas-metrics-coach skill. The financial-analyst skill provides a 5-phase workflow for ratio analysis, DCF valuation, budget variance analysis, and rolling forecasts, supported by Python scripts (ratio\_calculator.py, dcf\_valuation.py, budget\_variance\_analyzer.py, forecast\_builder.py) that use only the Python standard library. The saas-metrics-coach skill focuses on SaaS operating metrics like ARR/MRR, churn, CAC/LTV, and NRR. The router directs users to the appropriate skill based on their request, and both skills include reference guides, templates, and sample data to support analysis.

finance/skills · high confidence

A new \marketing/landing\ skill has been added, providing a \/cs:landing\ command that generates a single, self-contained HTML landing page file. Unlike the existing \product-team/skills/landing-page-generator\ which targets conversion-optimized lead generation with Next.js TSX, this skill focuses on visual-premium one-pagers featuring GSAP 3D animations, scroll-triggered reveals, and mouse-parallax depth. The skill enforces a structured intake process (product pitch, audience, brand colors, tone) and outputs a single HTML file with all CSS/JS inline (except Google Fonts and GSAP CDN), ensuring zero build steps and immediate deployability to any static host.

marketing · high confidence

The litreview skill has been introduced to generate structured literature orientation guides (DOCX) rather than finished reviews. It defaults to a free, keyless search lane using PubMed E-utilities and OpenAlex, with an optional Consensus MCP enhancement lane when available. The skill guides users through a structured intake (research question, framework selection via PICO/SPIDER/Decomposition, and depth tier), performs targeted searches, and synthesizes findings into an 8-section guide including priority reading lists, field timelines, and key research groups. Supporting scripts handle citation tracking, cross-search intelligence (repeat-hit and recurring-author analysis), and framework recommendation heuristics.

research/litreview/skills · high confidence

Added four new agent configuration files in the marketing domain: \cs-aeo\ (Answer Engine Optimization specialist), \cs-content-creator\ (long-form content production with quality gates), \cs-demand-gen-specialist\ (acquisition funnel and paid ads), and \cs-webinar-marketer\ (webinar funnel planning and rescue). These files define the agents' purposes, voice, skill integrations, and specific Python tool workflows for the marketing area.

agents/marketing · high confidence

Added three new marketing skills to the skills directory: \ab-test-setup\ provides a structured workflow for designing A/B tests, including a bundled \sample\_size\_calculator.py\ script for determining statistical power and test duration; \ad-creative\ assists in generating and iterating ad copy across platforms like Google, Meta, and LinkedIn, supported by an \ad\_copy\_validator.py\ script that checks character limits and flags platform-specific rejection triggers; and \aeo\ (Answer Engine Optimization) helps optimize content for citation by AI models, featuring an E-E-A-T auditing methodology and citation tracking tools. These skills introduce new CLI utilities and reference documentation to support experimentation, paid advertising production, and AI-first content strategy.

marketing-skill/skills · high confidence

A new \patent\ skill has been added to the research pack, enabling users to conduct structured patent searches for novelty, freedom-to-operate (FTO), competitive landscape, acquisition diligence, and litigation prior-art. The skill enforces a mandatory intake process where users must select one of five specific sub-use-cases, which then dictates the search strategy, ranking heuristics, and output emphasis. It integrates with Google Patents, Espacenet, USPTO, and optionally Lens.org (via BYOK) to perform multi-source searches with CPC/IPC class follow-ups and family resolution to deduplicate cross-jurisdiction filings. The output is an editable Word document (.docx) containing a verdict, ranked closest art with claim-text extraction, and a full audit log, with mandatory legal disclaimers for novelty and FTO analyses.

research/patent, research/syllabus · high confidence

The agents/product directory now includes definitions for five specialized agents: cs-product-manager (orchestrating feature prioritization and PRD development), cs-product-strategist (handling OKR planning and competitive analysis), cs-product-analyst (managing KPIs and experiment design), cs-agile-product-owner (focusing on sprint planning and backlog refinement), and cs-ux-researcher (covering research planning and persona generation). These markdown files define the agents' purposes, integrated skills, Python tools, and workflows, enabling the system to delegate specific product management tasks to these specialized agents.

agents/product · high confidence

Added three new skills to the product-team skills library: Competitive Teardown, which provides a structured workflow, scoring rubric, and Python-based matrix builder for analyzing competitors; Experiment Designer, which guides A/B test planning with statistical guardrails and includes a sample size calculator script; and Landing Page Generator, which produces high-converting Next.js/React components with Tailwind CSS, complete with design style references, copy frameworks, and SEO checklists.

product-team/skills · high confidence

Introduces a new handoff skill for the productivity area that compacts conversation context into a structured document for a fresh agent. The skill includes first-run setup to configure save locations (OS temp, home, or project-local), a mandatory checklist for summarization, and a redaction linter to prevent secrets leakage. It features SessionStart and SessionEnd hooks to auto-load previous handoffs and remind users to create one, along with slash commands \/cs:handoff\ and \/cs:handoff-setup\.

productivity/handoff · high confidence

A new PreToolUse hook has been added to the security-guidance plugin that intercepts Edit, Write, and MultiEdit operations to warn and block 12 common security anti-patterns before they are committed. The hook detects command injection (exec, os.system, subprocess shell=True), XSS (innerHTML, dangerouslySetInnerHTML, document.write), SQL injection (f-string/.format), unsafe deserialization (pickle, yaml.unsafe\_load), code injection (eval, new Function), and GitHub Actions workflow injection. It uses session-state caching to prevent duplicate warnings and persists debug logs to \~/.claude/security-warnings-log.txt. The hook can be disabled per-session using the ENABLE\_SECURITY\_REMINDER=0 environment variable.

engineering/security-guidance · high confidence

This release adds a comprehensive set of new slash commands and their corresponding documentation files to the \commands/\ directory. Engineering workflows now include \/a11y-audit\ for WCAG 2.2 accessibility scanning, \/chaos-experiment\ for resilience testing, \/karpathy-check\ for code quality reviews, \/operator-audit\ for Kubernetes operator validation, \/flag-cleanup\ for feature flag hygiene, \/focused-fix\ for systematic module repair, \/plugin-audit\ for skill/plugin validation, \/pipeline\ for CI/CD generation, and \/changelog\ for git history management. Marketing capabilities are expanded with \/cs:aeo\ for Answer Engine Optimization, \/cs:webinar\ for webinar funnel management, and \/competitive-matrix\ for competitive analysis. Product and finance tools include \/code-to-prd\ for reverse-engineering codebases into PRDs, \/financial-health\ for financial analysis, \/okr\ for OKR generation, and \/persona\ for user persona creation. Additionally, a suite of engineering review commands (\/cs:fullstack-review\, \/cs:backend-review\, \/cs:frontend-review\, \/cs:engineer-grill\) has been added to guide architectural decisions using forcing questions and specialist forking. A \google-workspace\ command is also introduced for Google Workspace CLI administration.

commands · high confidence

A new standalone plugin, \general-counsel-advisor\, is introduced to provide startup-focused legal triage for contract risk scanning, IP strategy, term sheet analysis, and regulatory landscape mapping. It is dual-published, meaning the same content is also bundled within the \c-level-skills\ aggregate, with a sync script keeping the two locations in sync. The plugin includes a \plugin.json\ manifest (version 2.9.0) and a skill definition that leverages two Python scripts: \contract\_risk\_scanner.py\ to identify common founder-killer clauses in agreements, and \term\_sheet\_analyzer.py\ to score term sheets on founder-friendliness across 12 dimensions. The skill documentation and reference materials cover standard startup contracts (MSA, SaaS, NDA, DPA, employment, contractor, equity), IP protection strategies, and regulatory triggers (HIPAA, GDPR, FDA, fintech, etc.), explicitly disclaiming that it is not a substitute for licensed counsel.

c-level-advisor/general-counsel-advisor · high confidence

A new \weekly-review\ plugin (v2.11.2) has been added to the productivity area to help users close open loops and maintain a trusted personal system. It implements David Allen's three-phase GTD ritual—GET CLEAR, GET CURRENT, and GET CREATIVE—using three deterministic Python scripts: an open-loop scanner to inventory unchecked checkboxes and stale files, a checklist gate that refuses a COMPLETE verdict if any mandatory GET CURRENT step is skipped, and a commitment auditor that scores project health (0-100) and flags stalled or actionless items. The plugin is accessible via the \/cs:weekly-review\ command or natural language triggers like "run my weekly review," and is distinct from the existing \capture\ (intake) and \reflect\ (per-conversation) plugins.

productivity/weekly-review · high confidence

Adds a new \workflow-builder\ skill that enables users to design, scaffold, and validate deterministic multi-agent workflow scripts (.js) for Claude Code's Workflow tool. The skill includes an intake recommendation engine to infer topology from vague requests, a validator to enforce hard rules (pure-literal meta, no non-determinism, guarded loops), and a scaffolder to generate starters for five topologies (fan-out, pipeline, barrier, loop, judge-panel). It also provides persona agents, slash commands, and reference documentation to guide users through the workflow authoring process.

engineering/workflow-builder · high confidence

The Playwright Pro plugin now includes shell hooks that automatically detect Playwright configurations at session start and validate test files after edits. When a session begins, the system checks for common Playwright config files and reports the number of test files found. After writing or editing test files, the plugin runs a validation check that warns against common anti-patterns such as using waitForTimeout, non-web-first assertions, hardcoded localhost URLs, or deprecated page.$() methods, helping users write more robust Playwright tests.

engineering-team/playwright-pro/hooks · high confidence

The \engineering/universal-scraping-architect\ location now provides a complete, standalone skill for designing robust data-extraction pipelines. It introduces a mandatory validation gate—enforced by \scripts/validate\_extraction.py\—that prevents unvalidated or malformed data from being delivered to the user. The skill supports three intelligent routing modes: Mode 1 uses the Firecrawl API for public, JS-heavy, or bulk crawling tasks; Mode 2 uses local Python (BeautifulSoup, pandas) for static HTML or sensitive local files; and Mode 3 combines both for hybrid workflows. The package includes runner templates (\firecrawl\_example.py\, \local\_bs4\_example.py\) that enforce safety standards like \robots.txt\ compliance, token-budget tracking, and BYOK API key handling, alongside reference guides for technical integration and ethical scraping practices.

engineering/universal-scraping-architect · high confidence

The skill-doctor plugin has been rebuilt from the upstream warpdotdev version to enforce a stricter, evidence-based grading workflow. A new deterministic aggregation gate (score\_aggregator.py) now owns all scoring arithmetic, rejecting fabricated scores, unsampled sessions, and suggestions that lack a cited session ID or diff. Privacy is hardened with mandatory secret redaction (12 patterns) before any transcript touches disk, restrictive file permissions (0600/0700), and bounded session file reads. The report renderer is now a self-contained, zero-JS HTML page with dark-mode support, and the tool no longer relies on Warp-specific session sources, targeting Claude Code and Codex sessions instead.

engineering/skill-doctor · high confidence

Introduced package.json files for the new BrowserStack and TestRail MCP integrations within the Playwright Pro engineering team, establishing their runtime and development dependencies. Added a sample web application package.json to the dependency-auditor skill to serve as a test subject for auditing capabilities. Defined runtime dependencies (firecrawl, pandas, requests, beautifulsoup4) for the Universal Scraping Architect skill and configured Python testing infrastructure (pytest) via pyproject.toml and requirements-dev.txt.

(dependencies) · high confidence

Written by watchdog.canine.dev from the codebase's own history, inside the signed delivery this page is composed from.

• First survey — no prior run to compare against. CAI 47.

• Code Health 39
• Architecture 100
• Maturity 79
• Readiness 50
• Security 72
• Accessibility 43

• 300 commits — 210 feature/other, 90 fixes

• (repo) — 73 commits
• engineering/agent-memory — 47 commits
• (root) — 16 commits
• .codex/skills — 16 commits
• engineering/skillopt-sleep — 12 commits
• engineering/deep-learning-book — 11 commits
• c-level-advisor/arquiteto-de-empresa — 9 commits
• engineering/human-gate — 9 commits
• engineering/book-to-skill — 8 commits
• engineering-team/playwright-pro — 7 commits
• engineering/memory-engineering — 6 commits
• productivity/meetings — 6 commits
• scripts/derive_counters.py — 6 commits
• .claude-plugin/marketplace.json — 5 commits
• productivity/fable-goal — 5 commits
• .codex/skills-index.json — 4 commits
• .github/workflows — 4 commits
• agent-launcher/skills — 4 commits
• marketing-skill/skills — 4 commits
• productivity/deep-work — 4 commits

• fix: Fix Star History Chart link in README
• fix: Merge pull request #917 from Jrtorres13/fix/c-level-advisor-readme-links
• fix: Merge pull request #929 from warnes/fix/gws-recipe-runner-subprocess-hardening-v2
• fix: Merge pull request #936 from benrfairless/fix/frontmatter-yaml-validation
• fix: Merge pull request #937 from benrfairless/fix/validator-and-model-freshness
• fix: Merge pull request #938 from benrfairless/fix/stale-model-references
• fix: Merge pull request #964 from automotua/fix/copilot-cli-description-limit-dev
• fix: fix(agent-launcher): make SKILL.md path references resolvable — unblocks the G1 CI gate for every PR
• fix: fix(agent-launcher): resolve gate-G1 phantom-path findings in the 6 SKILL.md files
• fix: fix(agent-memory): address automated review; harden the session-id fallback
• fix: fix(agent-memory): checker's paths broke on its own documented move — silently
• fix: fix(agent-memory): example violated additionalProperties — and the checker hid it
• fix: fix(book-to-skill): close the workdir race with fd pinning
• fix: fix(book-to-skill): correct emitter docstring drift and the resident-core figure
• fix: fix(book-to-skill): guard EPUB XML, cap zip expansion, refuse symlinked trees
• fix: fix(book-to-skill): private per-invocation workdir; share budget constants
• fix: fix(book-to-skill): route the sniff path through the zip budget; emitter fixes
• fix: fix(book-to-skill): satisfy CI path, smoke, and stdlib-shadowing gates
• fix: fix(book-to-skill): skill-quality audit — runnable docs, honest gates
• fix: fix(ci): add skillopt_sleep internal modules to smoke_exceptions.txt
• …and 280 more

• 0 containers · 1 bounded contexts · 0 dependency edges (baseline)

Written by watchdog.canine.dev from the codebase's own history, inside the signed delivery this page is composed from.

alirezarezvani/claude-skills 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.