菜单
Verify the framework's toolchain (git, bash, jq, gh, yq, PowerShell 7+, Node.js/npx, uv/uvx, shellcheck, Claude Code CLI) is installed and working — use when setting up the framework, after system updates, or when hooks, validators, MCP servers, or the executor agents fail with "command not found".
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Guide for adding a new real Claude Code hook to this framework as a .ps1/.sh pair (registration in hooks/hooks.json as a shell-form chain, tests in both suites, validation) — use when asked to create, generate, or extend a hook or quality gate.
Recommends which framework agent (or multi-agent workflow) should handle a task; use when the user asks which agent to use, how to delegate a piece of work, or how to sequence agents for a multi-domain task.
Diagnose and fix agent routing, loading, and configuration issues by inspecting agents/*.md frontmatter, the claude.json registry, hook registration, and validator output — use when an agent is "not found", tasks route to the wrong specialist, models mismatch, hooks do not fire, or validate-consistency.sh fails.
Git branching, commit, PR, conflict-resolution, and release guidance integrated with the framework's peer-review final gate — use when planning branches, writing commit messages, preparing or opening PRs, resolving merge conflicts, cutting releases, or when the Stop gate blocks a session over unreviewed committed work.
Iteratively improve a code change by having the matching specialist agent score the diff 0-100 against an explicit rubric, rewriting the weakest parts, and repeating until the score plateaus. Use after the quality bar is green (build/lint/format/tests clean) but before the review gates — e.g. polishing an implementation or refactor, or when asked to "clean up" code without concrete criteria. Code-side sibling of self-scoring-loop.
Iteratively improve a non-code deliverable (spec, plan, document, design) by writing an explicit rubric, scoring the work 0-100 against it, rewriting the weakest parts, and repeating until the score plateaus. Use when a deliverable has a quality bar but no test suite — e.g. before handing over a spec, proposal, architecture doc, or report.
| name | dependency-checker |
| description | Verify the framework's toolchain (git, bash, jq, gh, yq, PowerShell 7+, Node.js/npx, uv/uvx, shellcheck, Claude Code CLI) is installed and working — use when setting up the framework, after system updates, or when hooks, validators, MCP servers, or the executor agents fail with "command not found". |
Verify that every external tool this framework actually depends on is installed, on PATH, and at a workable version. Run the probes below, report a concise status table, and give platform-specific remediation for anything missing. Never invent results — run each probe and report exactly what it returns.
/agentic-framework:validate-hooks or /agentic-framework:analyze-framework report tool-related failures| Tool | Required by |
|---|---|
| git | Hard requirement; the Stop-gate hook's branch-vs-base checks; all repo workflows |
| bash (Git Bash on Windows) | scripts/validate-consistency.sh, scripts/validate-hooks.sh, scripts/validate-framework.sh, scripts/generate-docs.sh, tests/consistency.test.sh |
| jq | The bash validators (validate-consistency.sh aborts without it); parsing the registry (claude.json) and config files |
| gh (GitHub CLI) | The command-line executor agents (bash-expert, powershell-expert): PR/issue/run queries, log grinding, gh api reads; must be authenticated |
| yq (mikefarah v4) | The executor agents: YAML processing and agent-frontmatter extraction (yq --front-matter=extract) |
PowerShell 7+ (pwsh) | Windows: installer scripts require pwsh 7, and the hook chain falls back to it (hooks/dispatch.sh runs the POSIX .sh directly when jq is on PATH in Git Bash; .ps1 via pwsh otherwise). Linux/macOS: optional — hooks run as POSIX shell; pwsh only needed for the optional .ps1 test suite (tests/hooks.test.ps1). |
| Node.js + npx | The filesystem, context7, and sequential-thinking MCP servers (launched via npx -y) — only needed if the optional agentic-framework-mcp plugin is installed |
uv (uvx) | The serena and fetch MCP servers — only needed if the optional agentic-framework-mcp plugin is installed |
| Claude Code CLI | Host runtime that loads the agentic-framework plugin and executes hooks |
| shellcheck (optional) | Linting scripts/*.sh; nice-to-have, nothing hard-fails without it |
Run each probe and classify the result as OK (with version), MISSING, or WRONG VERSION. On Windows, run PowerShell probes in pwsh and shell probes in Git Bash.
git --versiongit -C <repo> rev-parse --is-inside-work-tree prints true.bash --version (first line). On Windows, confirm Git Bash specifically: bash -c 'echo ok' must print ok; Git for Windows installs it at C:\Program Files\Git\bin\bash.exe.jq --versionjq -e '.sub_agents | length > 0' claude.json from the repo root exits 0.gh --version, then gh auth status (exit 0 = authenticated; never pass --show-token).gh auth login.yq --versionhttps://github.com/mikefarah/yq v4+. The Python yq (a jq wrapper) has incompatible syntax and does NOT satisfy this dependency.pwsh -NoProfile -Command '$PSVersionTable.PSVersion.ToString()'powershell.exe (Windows PowerShell 5.1) is NOT sufficient — the hook scripts refuse to run under it.|| arm runs it when sh is unavailable; with jq present, hooks run as POSIX .sh and skip pwsh entirely).node --version and npx --versionnpx -y <package>.uvx --version (or uv --version)claude --versionclaude mcp list should show filesystem, context7, serena, sequential-thinking, and fetch.shellcheck --versionVersion probes prove installation, not integration. When the user is troubleshooting (not just installing), also run:
pwsh -NoProfile -File tests/hooks.test.ps1 — exercises the .ps1 hook implementations end to end (needs pwsh 7+ and git). Windows: required check. Linux/macOS: optional verification.bash tests/hooks.test.sh — exercises the .sh hook implementations (needs bash and git). Linux/macOS: verifies POSIX hooks work. Windows: optional if Git Bash is available.bash scripts/validate-consistency.sh — runs the full consistency battery (needs bash and jq); it validates the registry (claude.json), agent frontmatter parity, and hook pair parity against hooks/hooks.json.bash scripts/validate-hooks.sh — focused hook pair-parity and dispatch check.If a smoke check fails while all version probes pass, the problem is configuration (PATH visible to Claude Code's hook shell, plugin installation state), not a missing tool — verify the agentic-framework plugin is installed and active: /plugin list should show it, and the session may need restarting for hooks to load.
winget install Git.Git (includes Git Bash)winget install jqlang.jqwinget install GitHub.cli, then gh auth loginwinget install MikeFarah.yqwinget install Microsoft.PowerShellwinget install OpenJS.NodeJS.LTSwinget install astral-sh.uv (or irm https://astral.sh/uv/install.ps1 | iex)winget install koalaman.shellchecknpm install -g @anthropic-ai/claude-code or the native installerbrew install git jq gh yq node uv shellcheckbrew install --cask powershell (provides pwsh)npm install -g @anthropic-ai/claude-code or the native installersudo apt-get install -y git jq shellcheck gh (gh may need the GitHub CLI apt repo); yq: download the mikefarah binary from GitHub releases (the apt yq is the incompatible Python wrapper)packages.microsoft.com), then sudo apt-get install -y powershellcurl -LsSf https://astral.sh/uv/install.sh | shnpm install -g @anthropic-ai/claude-code or the native installerPresent results as a single table: Tool | Status | Version found | Needed by | Fix (only for failures). Then a one-line verdict:
Do not fabricate percentages, scores, or tool inventories beyond the table above, and do not install anything without the user asking — report and recommend.