AGENT SKILL

setup

Report where verification artifacts land and write the repository's docs/conventions/verification.yaml settings. check (read-only) names the memory root the verification skills resolve and validates the settings file; apply writes the keys its arguments name, proof_level, live_workers or both, after confirmation. Use when: 'set up verification', 'configure the verification plugin', 'is verification configured', 'verification setup', 'where do verification manifests / baselines land', 'set the proof level for this repo', 'set live workers for this repo'. Actions: check (read-only, default) | apply [proof_level=VALUE] [live_workers=N]. Re-runnable.

License 许可
MIT
最近更新
2026-11-15
Stars
★ 30
Forks
7

01 前言 Front Matter

description
Report where verification artifacts land and write the repository's docs/conventions/verification.yaml settings. check (read-only) names the memory root the verification skills resolve and validates the settings file; apply writes the keys its arguments name, proof_level, live_workers or both, after confirmation. Use when: 'set up verification', 'configure the verification plugin', 'is verification configured', 'verification setup', 'where do verification manifests / baselines land', 'set the proof level for this repo', 'set live workers for this repo'. Actions: check (read-only, default) | apply [proof_level=VALUE] [live_workers=N]. Re-runnable.
argument-hint
[check|apply] [proof_level=<path|live|strict>] [live_workers=<n>]
user-invocable
true
disable-model-invocation
true

02 资源文件 2

03 SKILL.md 正文

Purpose

Artifact placement is fixed by the plugin's lifecycle artifact protocol (${CLAUDE_PLUGIN_ROOT}/reference/artifact-protocol.md): /verification:confirm writes its evidence manifest and /verification:measure its baselines and raw captures into the memory slice <memory_dir>/<slug>/, never committed. <memory_dir> is .work/ unless the project's own instructions declare another root. Nothing about placement is configured.

The plugin owns one tracked settings file, docs/conventions/verification.yaml, the repository layer of its settings, proof_level and live_workers=<n> (values, layers and the level rules: ${CLAUDE_PLUGIN_ROOT}/reference/config.md; schema: ${CLAUDE_PLUGIN_ROOT}/schemas/verification.schema.json). check reports it; apply writes it only when the operator names the key or asks to set the proof level. Idempotent: re-running reads the current state again.

check (read-only)

Report a PASS/INFO/WARN table. Do not write anything.

  1. Memory root. Look for a working-docs root declared in the repo's own CLAUDE.md, AGENTS.md, or .claude/rules. Report the declared root, or .work/ when none is declared, as INFO.
  2. Ignore state. Run git check-ignore -v <memory_dir>/probe/baselines/x on a representative file path (a bare directory misses ** patterns). A match is PASS and names the rule. No match is INFO: the memory root's own self-ignoring .gitignore is created by the first memory-slice write, announced, so an absent guard before any run is expected.
  3. Settings file. Run node "${CLAUDE_SKILL_DIR}/scripts/setup-apply.mjs" --check from the project root and add each line it prints with its own prefix: INFO when the file is absent (proof_level and live_workers come from userConfig or their defaults), PASS with each key's value when it validates, WARN when it does not (a proof_level outside path, live, strict, a live_workers that is not a whole number of at least 1, a key set twice, an empty value, null, a key the schema does not list, a file that does not parse, a list or map, a symlink or hard link), quoting the file, key and value. A key the schema does not list is ignored: the other keys keep their PASS lines, so a valid proof_level floor beside it counts. A WARN never stops /verification:confirm: it names the value, drops the layer and resolves that key's default (path, 1). apply is the fix. /verification:confirm reads the default branch's committed copy, so say that a working-tree value takes effect once it is merged there.

apply

  1. Resolve the values. Take each key the arguments name, proof_level, live_workers or both, and use a complete <key>=<value> as given. For each named key without a value, ask once, recommendation first:

    • proof_level: path, the default, unless the team wants every runnable change driven live (live) or per-phase unit, live and performance proof (strict). Before recommending live or strict, say that a member on an older verification release does not read this file, so the level only holds once every member has upgraded.
    • live_workers: 1, the default, unless the team wants the live drive split across workers by feature-map entry point.

    Never invent a key the schema does not list, and never adjust a value: the script decides whether it is allowed.

  2. Write. One call carrying every resolved key, each <key>=<value> as one quoted argument:

    node "${CLAUDE_SKILL_DIR}/scripts/setup-apply.mjs" "proof_level=live"
    node "${CLAUDE_SKILL_DIR}/scripts/setup-apply.mjs" "live_workers=3"
    node "${CLAUDE_SKILL_DIR}/scripts/setup-apply.mjs" "proof_level=live" "live_workers=3"
    

    Pass a key the operator accepts at the plugin default as --default <key> instead of <key>=<value>, so a later change to that default reaches it; <key>=<value> records the team's choice, which the script never changes. The script also records written_against and provenance:, so a file with no record, or one written against another plugin version, shows a diff even when no value changes.

    The script checks each value against the schema (live_workers takes an unquoted whole number of at least 1, so 0, 1.5 and '2' are refused), refuses a key given twice, validates the existing file, and validates the whole resulting document before it writes. In the existing file it overwrites only a value outside the key's values, an empty value or null; it refuses a key set twice, a map or list in block or flow form, an empty quoted string, an unknown key set twice or holding a map or list, or a file that does not parse; any other unknown key is named in one warning and its line kept as written. It writes only <git toplevel>/docs/conventions/verification.yaml. It resolves a symlinked root first, then refuses a root that is $HOME or an ancestor of it, a symlinked docs, docs/conventions or target, a hard-linked target, or a docs/conventions that resolves outside the repository, checked again before each directory it creates and right before the write. Every refusal is one line, exits 1 and leaves the file as it was; a failed write removes only its own temp file. A missing file is created; a value already in place prints already configured.

  3. An existing file that would change exits 3 and prints a unified diff without writing. Show the operator that diff and ask whether to write it. Only on an explicit yes, re-run the same call with --yes; on anything else, stop with the file unchanged. When the convention-home offer prints below the diff, ask about it too (the next step) before any re-run, since --yes writes both.

  4. The convention-home offer. When the repository's AGENTS.md or CLAUDE.md binds no convention home, the script prints an offer: with the operator's confirmation it adds a convention-home pointer region to that instruction file (creating AGENTS.md when neither exists or CLAUDE.md only imports it) and the personal override ignore line docs/conventions/**/*.local.* to .gitignore, the only files apply writes outside docs/conventions/. It prints the offer at the end of an exit 0 run and below the diff of an exit 3 run, so show the operator stdout on exit 0 too. Ask whether to add the region and the ignore line. Only on an explicit yes, re-run the same call with --yes, which also writes any diff shown with it. On a no, run node "${CLAUDE_SKILL_DIR}/scripts/setup-apply.mjs" --decline alone: it records the decline in the instruction file, and the next run does not offer again. To write a shown diff without the region, run --decline first, then the same call with --yes. When the instruction file or .gitignore is a symlink or hard link the script will not write through, it prints a one-line refusal in place of the offer, with the same exit code: show it, and do not re-run with --yes, which that refusal stops with exit 1. A refused instruction file refuses --decline too, so the operator fixes the link or adds the region by hand.

  5. Verify. Re-run the settings-file check row and report the value it prints, not the value written. Then run git check-ignore -v docs/conventions/verification.yaml (a match is FAIL with the pattern) and git ls-files --error-unmatch docs/conventions/verification.yaml (non-zero right after a fresh write means "written but untracked: commit it to share with the team").

Output

The check table. After apply: the diff shown, the written path, the value check reads back for each key written, and whether the file is tracked.

Next

/verification:confirm, which resolves proof_level from this file on the default branch.

What this skill does NOT do

  • Run a verification pass. That is the plugin's verification skills (/verification:confirm, /verification:measure).
  • Write any settings file other than docs/conventions/verification.yaml, edit any ignore file beyond the personal override line the operator confirms in the convention-home offer, or write the plugin directory or the plugin data directory (${CLAUDE_PLUGIN_DATA} is for caches and generated state only).
  • Write Claude Code user settings or pluginConfigs. The per-user proof_level and live_workers options are set through /plugin configure, not here.

04 同仓库更多 Skills

melodic-software
melodic-software/claude-code-plugins

track

Backlog reads through the bound tracker: search open and closed items, view one item with its comments and state history. Use when another skill searches or reads tracker items.

★ 30
melodic-software
melodic-software/claude-code-plugins

reviewer-media

Reviewer media for one pull request: decides whether it needs a narrated walkthrough, builds the storyboard from /review:explain-change's digest, has /explainer-video:produce render it and hands the manifest to /source-control:pull-request media. Use when asked for 'reviewer media', 'a walkthrough video for this PR', 'a produced video of this change', or when pull-request ready or monitor hands a PR over.

★ 30
melodic-software
melodic-software/claude-code-plugins

route

Model and effort per agent role (orchestrator, worker, verifier, retrieval) for a multi-agent run, resolved through the bundled defaults and the user, team and personal override config layers into fenced JSON with single-agent and fan-out variants and the layer behind each value. Use when asked 'which model for the subagents', 'route the roles' or 'model routing', or before a skill passes a role's model on a spawn.

★ 30
melodic-software
melodic-software/claude-code-plugins

route

Model and effort per agent role (orchestrator, worker, verifier, retrieval) for a multi-agent run, resolved through the bundled defaults and the user, team and personal override config layers into fenced JSON with single-agent and fan-out variants and the layer behind each value. Use when asked 'which model for the subagents', 'route the roles' or 'model routing', or before a skill passes a role's model on a spawn.

★ 30
melodic-software
melodic-software/claude-code-plugins

principles

Module-design vocabulary for architecture findings, and the checks built on it. Use when: 'what is a deep module', 'what is a seam', 'deep vs shallow module', 'what is the deletion test', 'leverage and locality', 'module design vocabulary', 'which word do we use for X in architecture findings'. Scanning a codebase for friction is /architecture:improve.

★ 30
melodic-software
melodic-software/claude-code-plugins

write

Structured five-field bug report (title, steps to reproduce, expected vs actual, justified severity, suggested fix location) from an informal description, read-only. Use when the user names a defect they observed, or asks for one written up ('report a bug', 'file a bug', 'write this up as a bug').

★ 30