By kazukinagata
Re-verification probes for Claude Code plugin machinery (envs, substitution, hooks, userConfig, slash vs natural, blocks, Cowork-specific behavior). Source: sandbox/research.md targeting v2.1.118-119; this plugin re-tests claims against the current Claude Code version.
Observation infrastructure canary. Verifies that plugin-level hooks fire (via session-start log), that skill frontmatter hooks register and fire after first invoke, and that bash tool subprocess can write to the findings directory. Run this BEFORE any other probe — if this FAILs, the verdict of every later probe is CANARY-FAILED.
Probe env propagation across 3 process tiers (§1.1). Compares which CLAUDE_* env vars are set in plugin-level hook vs skill frontmatter hook vs Bash tool subprocess.
Probe ${VAR} substitution allowlist in skill body (§1.2). Tests CLAUDE_PLUGIN_ROOT/DATA/SKILL_DIR/SESSION_ID/PROJECT_DIR substitution at invoke time.
Probe that hook commands run under /bin/sh (dash on WSL Ubuntu), not bash (§1.3). Bash-specific constructs should produce Bad substitution / syntax errors.
Confirm that userConfig.api_secret (sensitive=true) is exposed as plain text via CLAUDE_PLUGIN_OPTION_API_SECRET in plugin-level hook env (§1.4). "sensitive" affects storage only, not runtime exposure.
Executes bash commands
Hook triggers when Bash tool is used
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
This plugin requires configuration values that are prompted when the plugin is enabled. Sensitive values are stored in your system keychain.
api_secretSensitive probe value (used for sensitive-leak probe). NOT a real API key.
${user_config.api_secret}hello_messageNon-sensitive probe value (stored in settings.json)
${user_config.hello_message}Claude Code plugin 機構の挙動を 22 個の probe skill で実機検証し、公式ドキュメントには書かれていない仕様や、ドキュメントと挙動が矛盾する箇所を洗い出したレポジトリ。
検証対象:
docs/team-report.md に 約 2,400 行の落とし穴まとめがあります。主なトピック:
CLAUDE_PLUGIN_ROOT / CLAUDE_PLUGIN_DATA / CLAUDE_PROJECT_DIR 等が plugin-level hook / skill frontmatter hook / Bash tool subprocess の 3 階層で非対称に渡される${VAR} 事前置換の tier 別 allowlist:skill body は最も広い、skill frontmatter は ${CLAUDE_PLUGIN_ROOT} のみsensitive: true userConfig の挙動:plugin-level hook には平文で env 渡し、skill body は Claude Code 本体が block 文字列に置換して leak 防止PreToolUse:Skill は自然言語のみ、UserPromptExpansion は slash のみclaude_code.skill_activated event による 3 経路 (user-slash / claude-proactive / nested-skill) の追跡方法verifier/ # 22 probe skill を持つ検証用 plugin 本体
verifier-violator/ # validator block を意図的にトリガーする違反 plugin
docs/
team-report.md # 全 finding をまとめた本文(一番読むべき)
check-matrix.md # probe 別 verdict 表
cowork-runbook.md # Cowork 実機検証手順
methodology.md # 検証手法
scripts/ # log 解析 / assert / Cowork zip 生成
findings/ # 実行結果(per-version observations.md / report.md のみ git 管理)
MASTER-RUNBOOK.md # 22 probe を順に回すための手動手順
LICENSE
claude --print 等の自動化は採用しない。人間が claude 対話セッションを起動し、/verifier:NN-... を順に叩くscripts/assert.sh)のみ~/.claude/ を汚さないよう CLAUDE_CONFIG_DIR=$(pwd)/findings/claude-home/ で project-local に閉じ込めるfindings/v<version>/report.md に verdict 6+1 種(PASS / FAIL / DOC-ALIGNED / PARTIAL / UNKNOWN / CANARY-FAILED / MANUAL-OK)で残すcd /path/to/analyzing-claude-plugin
. scripts/_env.sh
./scripts/capture-cli-help.sh
# findings/cli-help/*.log と STEP0-SUMMARY.md を生成
. scripts/_env.sh
bash scripts/install-marketplace.sh
別 terminal で:
. scripts/_env.sh
claude
claude prompt 内で:
/verifier:00-canary
完了後に exit して:
./scripts/assert.sh 00
# PASS が返れば観測基盤 OK
MASTER-RUNBOOK.md を順に追って 22 probe を回す。各 probe で assert.sh NN を叩いて verdict を確認。
最後に ./scripts/assert-all.sh で findings/v<version>/report.md を生成。
./scripts/package-cowork.sh
で zip を作成、docs/cowork-runbook.md の手順で Claude Desktop にアップロード。
| トピック | 観測 | 関連 § |
|---|---|---|
${VAR} 事前置換 allowlist | tier ごとに別 allowlist で運用、skill frontmatter は ${CLAUDE_PLUGIN_ROOT} のみ | team-report §1.2 |
sensitive: true の skill body block | Claude Code 本体が [sensitive option 'KEY' not available in skill content] に置換して leak を防ぐ | team-report §1.4 |
| Cowork architecture | host-adjacent VM + virtio-fs。plugin-level hook は host 側、Bash tool は cloud VM 側 | team-report §2.0 |
| Cowork validator | CLI より厳しい:kebab-case 強制、description 内の ${...}/<...> 拒否、UserPromptExpansion event 拒否 | team-report §2.16 |
OTel skill_activated | invocation_trigger=user-slash/claude-proactive/nested-skill で 3 経路完全区別、Cowork でも emit | team-report Appendix B |
MIT License — LICENSE 参照。
確定申告を自動化する Claude Code Plugin。会社員+副業(事業所得・青色申告)の所得税・消費税確定申告をエンドツーエンドで支援。
An async task coordination protocol for autonomous AI agents. Agents discover, claim, execute, and hand off tasks through a shared board.
Notion provider for waggle. Enables team task management via Notion workspace.
SQLite provider for waggle. Local zero-config task management.
Turso provider for waggle. Remote SQLite for multi-agent coordination.
npx claudepluginhub kazukinagata/analyzing-claude-plugin --plugin verifierComprehensive skill pack with 66 specialized skills for full-stack developers: 12 language experts (Python, TypeScript, Go, Rust, C++, Swift, Kotlin, C#, PHP, Java, SQL, JavaScript), 10 backend frameworks, 6 frontend/mobile, plus infrastructure, DevOps, security, and testing. Features progressive disclosure architecture for 50% faster loading.
A growing collection of Claude-compatible academic workflow bundles. Covers scientific figures, manuscript writing and polishing, reviewer assessment, citation retrieval, data availability, paper reading, literature search, response letters, paper-to-PPTX conversion, and evidence-grounded Chinese invention patent drafting. Rules are organized as reusable skill folders with explicit workflows and quality checks.
Develop, test, build, and deploy Godot 4.x games with Claude Code. Includes GdUnit4 testing, web/desktop exports, CI/CD pipelines, and deployment to Vercel/GitHub Pages/itch.io.
Permanent coding companion for Claude Code — survives any update. MCP-based terminal pet with ASCII art, stats, reactions, and personality.
Comprehensive PR review agents specializing in comments, tests, error handling, type design, code quality, and code simplification
Comprehensive feature development workflow with specialized agents for codebase exploration, architecture design, and quality review