Skip to content
CAI
Software that uses CAICheck a score

anthropics/claude-plugins-official

32.7

Weak · 15 November 2026

43.7k

lines of production code

Python

with TypeScript, JavaScript

4

measurements over time

CAI band scale
CAI trend line
CAI lens gauges

What this system is

Features

Add fakechat channel plugin for local testing

A new 'fakechat' plugin has been added to external\_plugins/fakechat, providing a localhost web-based chat interface for testing the channel contract without requiring external messaging services. The plugin registers an MCP server exposing 'reply' and 'edit\_message' tools, allowing users to simulate user interactions and file uploads via a local UI served on port 8787 (configurable via FAKECHAT\_PORT). It is designed as a development tool for validating channel behavior, with no persistent history, access control, or external dependencies.

_external\plugins/fakechat · high confidence

Introduce Claude Security Plugin for static code analysis

Adds the Claude Security Plugin, a new capability that scans repositories for vulnerabilities and generates actionable reports. The plugin introduces a Python-based workflow (hooks, scripts, and libraries) that performs security scans in various modes (full repository, branch changes, or specific commits), records findings with CWE classifications, and renders results into SARIF 2.1.0 logs, JSONL, and Markdown reports. It also supports generating and applying suggested code patches for identified issues, integrating directly with the Claude Code environment to provide automated security reviews.

python · high confidence

Introduce Discord channel plugin for Claude Code

Adds a new Discord channel plugin that bridges Discord messages to Claude Code via an MCP server. The plugin enables users to connect a Discord bot to their assistant, supporting direct messages and guild channels with configurable access control (pairing, allowlist, or disabled policies). It exposes tools for the assistant to reply, react, edit messages, fetch history, and download attachments, while managing state (allowlists, pending pairings, channel settings) in a local JSON file.

_external\plugins/discord · high confidence

Introduce Telegram channel plugin with access control and MCP tools

Adds a new Telegram channel plugin that connects a Telegram bot to Claude Code via an MCP server. The plugin provides tools for the assistant to reply to, react to, and edit messages, and automatically downloads inbound photos. It includes a built-in access control system with a default pairing flow (users send a code to approve access), support for allowlists and group chats, and configurable delivery options like message chunking and reaction emojis. The plugin requires Bun to run and manages its state (tokens, access policies) in a configurable state directory.

_external\plugins/telegram · high confidence

Introduce receipts plugin for local usage impact reports

The new receipts plugin generates a personal impact report from local Claude Code session transcripts, helping users justify their usage or perform self-reviews. It mines \\~/.claude/projects\ for session data, cross-references local git history to count commits and PRs, and outputs a markdown report and a self-contained HTML receipt (with CSV export) to the user's home directory. The report tracks sessions, prompts, files and lines touched, and active days, broken down by project, while deliberately excluding dollar figures or per-activity spend breakdowns to avoid misleading precision. All processing is local with no network calls, ensuring privacy.

plugins/receipts · high confidence

Introducing Claude Security Plugin for in-session vulnerability scanning and patching

A new Claude Security plugin is available, bringing Anthropic's hosted security capabilities directly into your Claude Code session. It performs deep vulnerability scans of your codebase (or specific diffs) using a team of specialized agents that map architecture, hunt for flaws, and independently verify every finding before reporting. Scans produce human-readable reports, machine-readable JSONL, and SARIF logs, with automatic redaction of hardcoded credentials. For confirmed findings, the plugin can generate targeted patch files that are reviewed by independent agents to ensure they fix the issue without introducing new vulnerabilities or altering unrelated behavior, allowing you to apply fixes when you choose.

plugins/claude-security · high confidence

Introducing the code-modernization plugin

The new code-modernization plugin provides a guided workflow for modernizing legacy codebases, supporting same-stack version uplifts, rewrites into new technologies, and architectural rebuilds. It begins with a preflight check and assessment, then extracts business rules into testable specifications for human review. The plugin generates an approved modernization plan and executes the build (uplift, transform, or reimagine), followed by an independent verification step that compares old and new behavior to issue a PROVEN, PARTLY PROVEN, or NOT PROVEN verdict. It includes a live progress pane, a one-page offline report, and security hardening scans, while ensuring the original legacy source code remains untouched.

plugins/code-modernization · high confidence

New iMessage channel plugin for macOS

Adds a new external plugin that connects Claude Code to iMessage on macOS. The plugin reads the local Messages database (\chat.db\) for history and new messages, and sends replies via AppleScript. It includes built-in access control with a default allowlist policy (self-chat always allowed), support for group chats, and configuration via environment variables like \IMESSAGE\_ALLOW\_SMS\ and \IMESSAGE\_APPEND\_SIGNATURE\.

_external\plugins/imessage · high confidence

New math-olympiad plugin for competition math with adversarial verification

The new math-olympiad plugin enables solving competition math problems (IMO, Putnam, USAMO, AIME) using a multi-agent workflow that combines parallel solver attempts with fresh-context adversarial verification to catch errors that self-verification misses. The skill automatically triggers on keywords like 'IMO', 'Putnam', or 'olympiad', and uses calibrated abstention to avoid bluffing when confidence is low. It supports batch solving of full problem sets with parallel execution and produces clean, presentation-ready LaTeX/PDF output after verification passes.

plugins/math-olympiad · high confidence

New math-proof plugin with solo and siege skills for complex mathematics

The new math-proof plugin provides two skills for tackling hard, research-level mathematics problems, each producing a self-contained proof.md. The /math-proof:solo skill works through a problem in stages within a single session, using a notes file to preserve progress against output limits. The /math-proof:siege skill orchestrates a multi-agent process where a judge directs independent workers through multiple rounds to prove or disprove a goal, maintaining a ledger of results and requiring line-by-line verification before concluding. Both skills disable web tools and require specific environment settings for long responses and sub-agent management.

plugins/math-proof · high confidence

New project-artifact plugin for living project status pages

The new project-artifact plugin generates and publishes tabbed, self-contained HTML status pages for multi-workstream projects (such as migrations, launches, or research efforts) via Claude Code's built-in Artifact tool. These pages are published to a private claude.ai URL that can be shared with teammates and are designed to be 'living' artifacts: users can refresh them to re-gather live state (e.g., from GitHub PRs or trackers), redeploy to the same URL, and receive a delta summary of changes. The plugin supports domain-neutral structures (Overview, Workstreams, Attention, etc.) with specialized handling for software projects where workstreams are pull requests, and requires an interactive session with a claude.ai login to publish.

plugins/project-artifact · high confidence

New session-report plugin for analyzing Claude Code usage

The session-report plugin adds a new capability to generate self-contained HTML reports of Claude Code session activity. It includes an analyzer script that scans local transcript files to extract token usage, cache breaks, subagent activity, and skill invocations, and a styled HTML template that renders this data into an interactive dashboard with day-level timelines, drill-downs, and anomaly detection. This allows users to visualize and understand their coding session patterns and costs.

plugins/session-report · high confidence

New skill-creator plugin for building and evaluating skills

The new skill-creator plugin enables users to create, improve, and measure the performance of skills. It provides a structured workflow for drafting skills, running evaluations, and benchmarking performance with variance analysis. The plugin includes specialized agents for blind comparison of skill outputs, post-hoc analysis of why one skill outperformed another, and grading of execution transcripts against expectations. It also features an HTML-based review interface for evaluating test results and generating improvement suggestions.

plugins/skill-creator · high confidence

Behavioural changes

1 commit (0 fixes) modifying plugins/champion-kit

A change to existing behaviour in plugins/champion-kit — 1 commit, 13 files: “Add champion-kit plugin (\#7017)” (refs \#7017).

plugins/champion-kit · medium confidence · unverified

1 commit (0 fixes) modifying plugins/theme-examples

A change to existing behaviour in plugins/theme-examples — 1 commit, 23 files: “Add theme-examples: example backdrops for the Claude desktop app (\#7116)” (refs \#7116).

plugins/claude-for-manufacturing, plugins/theme-examples · medium confidence · unverified

141 commits (3 fixes) modifying plugins/security-guidance/hooks

A change to existing behaviour in plugins/security-guidance/hooks — 141 commits (3 fixs), 13 files: “security-guidance: 5 venv-specific err\_kind categories + stderr\_signature bucket (2.0.1 → 2.0.2)”; “security-guidance: follow the newest Opus instead of a pinned release (2.0.8 → 2.0.9)”; “security-guidance: count Windows sessions whose Python is behind a link”; and 137 more (refs \#1868, \#2099, \#2854, \#1358, \#1375, …).

plugins/security-guidance/hooks · medium confidence · unverified

3 commits (1 fix) modifying plugins/ralph-loop/scripts

A change to existing behaviour in plugins/ralph-loop/scripts — 3 commits (1 fix), 1 file: “Rename ralph-wiggum plugin to ralph-loop per legal guidance (\#142)”; “updates(staging): merge staging additions into main (\#677)”; “fix(ralph-loop): isolate loop state to the session that started it” (refs \#142, \#677).

plugins/ralph-loop/scripts · medium confidence · unverified

Fixes

3 commits (2 fixes) fixing plugins/hookify/hooks

A fix in plugins/hookify/hooks — 3 commits (2 fixs), 6 files.

plugins/hookify/hooks · medium confidence · unverified

5 commits (3 fixes) fixing plugins/ralph-loop/hooks

A fix in plugins/ralph-loop/hooks — 5 commits (3 fixs), 2 files: “address review: bound grep to tail -n 100; restore explicit error paths”; “Rename ralph-wiggum plugin to ralph-loop per legal guidance (\#142)”; “fix(ralph-loop): isolate loop state to the session that started it”; and 2 more (refs \#142).

plugins/ralph-loop/hooks · medium confidence · unverified

Fixes import path for rule engine configuration loader

The rule engine in the hookify plugin now correctly imports the Rule and Condition classes from the local core.config\_loader module, resolving a previous import error that prevented the plugin from loading.

plugins/hookify/core · high confidence

Test coverage

139 commits adding/updating tests in plugins/security-guidance/tests

Test coverage in plugins/security-guidance/tests — 139 commits, 72 files: “security-guidance: follow the newest Opus instead of a pinned release (2.0.8 → 2.0.9)”; “security-guidance: count Windows sessions whose Python is behind a link”; “security-guidance: open the review records and markers through one helper”; and 135 more (refs \#6537, \#6541, \#6379).

plugins/security-guidance/tests · high confidence · unverified

Dependencies

Add external channel plugins for Discord, Telegram, iMessage, and FakeChat

New external plugins have been added to enable communication channels for Discord, Telegram, iMessage, and a FakeChat test channel. Each plugin is packaged as a standalone Node.js module (using Bun) with its own package.json, depending on the @modelcontextprotocol/sdk and channel-specific libraries (discord.js, grammy, or zod). These plugins allow users to connect Claude to these messaging platforms via external server processes.

(dependencies) · high confidence

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

How this codebase got here

Baseline

  • First survey — no prior run to compare against. CAI 33.

Lenses

  • Code Health 37
  • Architecture 98
  • Maturity 76
  • Readiness 21
  • Security 64
  • Accessibility 30
  • Performance 60

Changes since last survey

  • 300 commits — 298 feature/other, 2 fixes

By area

  • (repo) — 112 commits
  • plugins/security-guidance — 108 commits
  • .claude-plugin/marketplace.json — 69 commits
  • .github/workflows — 3 commits
  • plugins/code-review — 2 commits
  • plugins/agent-sdk-dev — 1 commit
  • plugins/champion-kit — 1 commit
  • plugins/claude-for-manufacturing — 1 commit
  • plugins/kotlin-lsp — 1 commit
  • plugins/plugin-dev — 1 commit
  • plugins/theme-examples — 1 commit

Notable commits

  • fix: Merge pull request #6919 from anthropics/mhegazy/sg-add-failure-fix
  • fix: Merge pull request #6930 from anthropics/fix/sg-blank-base-url
  • change: Add Xcode plugin (xcode) (#6338)
  • change: Add champion-kit plugin (#7017)
  • change: Add claude-for-manufacturing plugin (#7084)
  • change: Add plugin.json to the 14 plugins that have none (#7013)
  • change: Add theme-examples: example backdrops for the Claude desktop app (#7116)
  • change: CI: validate in-repo plugin folders; point submissions to claude.ai/directory/manage (#7086)
  • change: Close every pull request opened by a non-member
  • change: Give ten plugins explicit versions
  • change: Keep git out of folders without a .git entry on older git versions
  • change: Keep the regex limit reset apart from the rest of session load
  • change: Keep whole fake secrets out of test sources
  • change: Merge branch 'main' into mhegazy/sg-baseline-subdir
  • change: Merge branch 'main' into mhegazy/sg-diff-seen-count
  • change: Merge branch 'main' into mhegazy/sg-gateway-login
  • change: Merge branch 'main' into mhegazy/sg-index-copy-gitlinks
  • change: Merge branch 'main' into mhegazy/sg-launcher-probe-timing
  • change: Merge branch 'main' into mhegazy/sg-network-failure-class
  • change: Merge branch 'main' into mhegazy/sg-pattern-fired-telemetry
  • …and 280 more

Architecture

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

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

Survey your own repository

anthropics/claude-plugins-official 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 15 November 2026 at a pinned commit. It is not a live figure and does not change until the project is measured again.
  • Measured at commit b5b01ac4f889fec0c4c6bb89f27ca7c8574473f6 — the exact code this score is about.
  • Scored under rubric-2026.11.4 — the same rubric and the same method as every other entry in this index.
  • Measured by Canine Code Analyzer (cca.canine.dev) using cca measure-2026.11.73+preprod-52de90b74216.