菜单
Use when starting vibe-validate work or picking a sub-skill. Router for the validate/fix loop, project setup, cache and locking, work recovery, extractor authoring, PR/CI monitoring, and non-Claude agent integration.
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
Use when monitoring PR/CI status after pushing, diagnosing CI failures, investigating flakes, or cleaning up merged branches. Triggers on watch-pr, CI checks, CI debug, PR status, merged-branch cleanup.
Use when validation fails but no errors are extracted, or when writing a custom error extractor for a tool vibe-validate does not recognize. Covers built-in extractors, vv create-extractor, and the plugin system.
Use when debugging cache misses, unexpected cache hits, or concurrency lock conflicts. Covers the git tree hash caching model, lock scopes, --no-lock, --no-wait, and the vv cleanup-temp maintenance utility.
Use when wiring vibe-validate into a non-Claude AI coding assistant (Cursor, Aider, Continue, or any other tool). Covers system-prompt additions, command wrappers, and output routing for agentic workflows.
Use when recovering lost work — accidentally deleted file, git restore wiped changes, failed rebase, editor crash, or branch switch lost untracked files. Covers vv history, vv snapshot, tree-hash recovery with git cat-file and git checkout.
Use when bootstrapping vibe-validate in a new repo, adjusting the config, or diagnosing setup. Covers vv init, vv doctor, vv config, templates, heterogeneous-language setup, generate-workflow, and dependency-lock-check.
| name | vibe-validate |
| description | Use when starting vibe-validate work or picking a sub-skill. Router for the validate/fix loop, project setup, cache and locking, work recovery, extractor authoring, PR/CI monitoring, and non-Claude agent integration. |
vibe-validate is a git-aware validation orchestration tool for LLM-assisted development. It runs your existing build/lint/test/typecheck commands, caches results by git tree hash for near-instant re-runs on unchanged trees, and reformats failures as structured YAML with stripped ANSI codes so an agent can read them cheaply.
This is a router skill. Load the sibling sub-skill that matches the work at hand — each sub-skill owns one slice of vibe-validate's surface and is designed to be pulled in on demand.
Good fits:
Poor fits:
npm test is sufficient| If you're working on... | Load |
|---|---|
The day-to-day validate/fix/revalidate loop — vv validate, vv pre-commit, vv state, vv run | vibe-validate:vv-validate-dev-loop |
Setting up vibe-validate in a repo — vv init, vv doctor, config schema, heterogeneous languages, dep-lock checks | vibe-validate:setting-up-projects |
| Debugging cache misses, cache hits that should've missed, or concurrency lock conflicts | vibe-validate:caching-and-locking |
| Recovering lost work from validation snapshots | vibe-validate:recovering-work |
| Validation fails but produces no extracted errors — custom extractor authoring | vibe-validate:authoring-extractors |
| Monitoring PR/CI status, debugging CI failures | vibe-validate:vv-watch-pr |
| Configuring vibe-validate for non-Claude tools (Cursor, Aider, Continue) | vibe-validate:integrating-agents |
vv --help # top-level help
vv validate # run validation, cached by tree hash
vv pre-commit # branch-synced validation (one checkpoint option)
vv state # query cached state (instant, no re-run)
vv run <cmd> # wrap any command for LLM error extraction
vv init # initialize vibe-validate.config.yaml
vv doctor # setup diagnostics
vv watch-pr # monitor PR/CI
vv history # validation history
vv snapshot # show current worktree snapshot
vv create-extractor # scaffold a custom extractor
Each sub-skill covers its slice of the CLI in depth — don't memorize this table, load the sub-skill when you need detail.
For a first-time adoption walkthrough, see adoption-guide.md in this directory. For full CLI syntax with every flag and exit code, see cli-reference.md in this directory.