[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M6HBVH8GX9JCMS5432JZVTMN":3,"skill-same:01M6HBVH8GX9JCMS5432JZVTMN":72},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":38,"linkOnly":39,"markdownContent":40,"markdownBody":41,"frontMatter":42,"refName":45,"commitSha":46,"skillDirectory":47,"forksCount":48,"repositoryHtmlUrl":49,"repositoryHomepageUrl":6,"ownerHtmlUrl":50,"ownerBlogUrl":6,"ownerDescription":51,"updatedAt":52,"install":53,"resources":65},"01M6HBVH8GX9JCMS5432JZVTMN","ship",null,"Run every canonical quality gate, push the branch, and open a pull request — release in one command. Use when: ship it, ready to ship, create the PR, push and open a PR.","SyloRei\u002Fclaude-godmode","SyloRei","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F5840195?v=4","01M6HBT6QSK7QM4DTVPNGARNP4","Andrey Prokofev","01M6HBT6WED1MEN8W7JYP412WE",3,"MIT","https:\u002F\u002Fgithub.com\u002FSyloRei\u002Fclaude-godmode\u002Fblob\u002Fmain\u002Fskills\u002Fship\u002FSKILL.md",[],[5,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37],"Bash","canonical","scripts","ship it","godmode","ready","read","command","branch","model","pr","open","push","pull","git","release","run","create","quality gate","2026-11-10 10:12:30",false,"---\nname: ship\ndescription: \"Run every canonical quality gate, push the branch, and open a pull request — release in one command. Use when: ship it, ready to ship, create the PR, push and open a PR.\"\nuser-invocable: true\ndisable-model-invocation: true\nallowed-tools: Read, Glob, Grep, Bash(git *), Bash(gh *), Bash(gm=*), Bash(*godmode-state*), Bash(*godmode-model*), Bash(*godmode-findings*), Bash(*\u002Fskills\u002Fship\u002Fscripts\u002Fgates.sh*), Bash(skills\u002Fship\u002Fscripts\u002Fgates.sh*)\n---\n\n# Ship\n\nTake a verified work unit to a pull request: run the canonical quality gates, then — only if every gate passes — push the branch and open a PR. This is the final step of the spine: `\u002Fmission` → `\u002Fbrief N` → `\u002Fplan N` → `\u002Fbuild N` → `\u002Fverify N` → **`\u002Fship`**.\n\n`\u002Fship` is side-effecting (it pushes and opens a PR), so it is **user-triggered only** (`disable-model-invocation: true`). It is never auto-invoked.\n\n---\n\n## Flags\n\n- **`--no-push`** (aliases: `--dry-run`, `--local`) — run every quality gate **and the findings gate (Step 1b)**, prepare git readiness (commit), then **stop before the push and the PR**. Draft and print the PR title\u002Fbody for review, but do **not** run `git push` or `gh pr create`, and do **not** record `status=shipped`. Use this to land a verified, committed, gate-green branch locally without publishing it. `--no-push` does **not** exempt the findings gate any more than it exempts the quality gates: open blocking findings BLOCK `--no-push` too (no commit), exactly as they BLOCK a normal ship.\n\nWhen `--no-push` is set the confirmation prompt is moot — there is nothing side-effecting to confirm — so skip it and run the gates and git readiness straight through.\n\n---\n\n## Confirm by default\n\n`\u002Fship` **confirms with the user before the side-effecting steps** (the push and the `gh pr create`). This confirm follows the shared convention in `rules\u002Fgodmode-recommend.md` (`godmode:recommend-convention`) — its confirm\u002Fproceed rendering: lead with a **Recommended: proceed** line carrying a visible one-line rationale, then let the user override.\n\nAfter the gates pass and the PR body is drafted, show the branch, target, and PR title\u002Fbody, then lead with the recommended-proceed confirm before pushing — for example:\n\n```\nRecommended: proceed — every quality gate and the findings gate are green, and\nthe PR body is drafted. Reply \"go\" to push and open the PR, or name what to change.\n```\n\nWait for explicit confirmation before pushing.\n\n**Exception — Auto Mode.** When `## Auto Mode Active` is present in context, skip the confirmation prompt: proceed straight to push and PR create on the default choices, and treat any user course-correction as normal input. The quality-gate block (Step 1) **and the findings gate (Step 1b)** are never skipped, in either mode — Auto Mode skips only the confirmation, never a BLOCK. (The findings gate's waive escape is inherently conversational, so it still requires an explicit user-named ID + reason even under Auto Mode; Auto Mode never auto-waives.)\n\n---\n\n**Resolving the godmode helpers.** The `godmode-*` helpers live in the plugin install dir — **not** the consumer repo you're working in — so a bare `bin\u002Fgodmode-*` path fails from another project's working directory. Every `bash` block below that calls one resolves its location into `$gm` first (plugin mode → `$CLAUDE_PLUGIN_ROOT\u002Fbin`, manual install → `~\u002F.claude\u002Fbin`, in-repo → `.\u002Fbin`) and calls `\"$gm\u002Fgodmode-\u003Cname>\"`. Keep the resolver line; never call a helper by a bare relative path.\n\n## Step 1: Quality gates (read from the canonical source)\n\nThe gates are defined in **`config\u002Fquality-gates.txt`** — one gate per line. **Read that file; do not hardcode the gate list here.** It is the single source of truth, so the gates stay in sync with the rest of the plugin.\n\nDon't re-derive the gate-running loop: invoke the bundled **`skills\u002Fship\u002Fscripts\u002Fgates.sh`**. It resolves `config\u002Fquality-gates.txt` across install modes (`${CLAUDE_PLUGIN_ROOT}\u002Fconfig` → `~\u002F.claude\u002Fconfig` → repo-relative), reads the gate list, prints one labelled status line per gate, and exits non-zero if any gate fails. You supply the per-gate verdicts via `GATES_RUNNER`: a command `gates.sh` invokes once per gate as `$GATES_RUNNER \"\u003Cgate text>\"` — exit 0 = pass, non-zero = fail. Wire `GATES_RUNNER` to your auto-detected per-gate dispatcher: for each gate text, run the project's matching command (typecheck, lint, test, build, secret scan; lint includes `shellcheck` clean for any `.sh` change) and exit with its status.\n\nResolve the script the same way across install modes, then run it:\n\n```bash\n# Locate the bundled gate runner — plugin mode, manual install, then repo-relative.\nGATES_SH=\"\"\nfor cand in \\\n  \"${CLAUDE_PLUGIN_ROOT:-}\u002Fskills\u002Fship\u002Fscripts\u002Fgates.sh\" \\\n  \"$HOME\u002F.claude\u002Fskills\u002Fship\u002Fscripts\u002Fgates.sh\" \\\n  \"skills\u002Fship\u002Fscripts\u002Fgates.sh\"; do\n  [ -n \"$cand\" ] && [ -f \"$cand\" ] && { GATES_SH=\"$cand\"; break; }\ndone\n[ -n \"$GATES_SH\" ] || { echo \"error: gates.sh not found\" >&2; exit 1; }\n\n# GATES_RUNNER is your auto-detected per-gate dispatcher: it receives one gate's\n# text as $1, runs the project command that gate maps to, and exits 0 (pass) or\n# non-zero (fail). gates.sh reads config\u002Fquality-gates.txt and calls it per gate.\n# GATES_RUNNER must resolve to a single executable — a script path or a command\n# name on PATH — NOT a shell function (invisible to the child process) and NOT a\n# space-separated string like \"bash runner.sh\" (invoked as one token, so it\n# would fail to find that file). Point it at an executable runner script.\nGATES_RUNNER=\"$run_one_gate\" bash \"$GATES_SH\"\n```\n\n`gates.sh` prints the report below; map each gate text to its detected command in your `GATES_RUNNER` dispatcher:\n\n```\nQuality gates (from config\u002Fquality-gates.txt):\n  [✓\u002F✗] Typecheck passes (zero errors)\n  [✓\u002F✗] Lint passes (zero errors; shellcheck clean for any .sh change)\n  [✓\u002F✗] All tests pass (existing + new)\n  [✓\u002F✗] No hardcoded secrets in the diff\n  [✓\u002F✗] No regressions in related functionality\n  [✓\u002F✗] Changes match the original requirements\n```\n\n**BLOCK on any failure.** If a gate fails: stop, report **which** gate failed and the error, and do **not** push or open a PR. Never `--no-verify`, never bypass a gate.\n\n- Test \u002F typecheck failures → use `\u002Fdebug`.\n- Lint → auto-fix where available, otherwise fix manually.\n- Multi-file fixes → spawn `@writer`.\n\n**Model profile.** Before spawning `@writer` (or any agent), resolve the active model profile from `${CLAUDE_PLUGIN_OPTION_MODEL_PROFILE:-balanced}`, then call the resolver `\"$gm\u002Fgodmode-model\" \u003Cagent>` (resolve `$gm` as the **Resolving the godmode helpers** note above shows) to obtain the model for that agent under the active profile. Pass that model to the Agent tool's `model` override at spawn time. The resolver also reports the agent's effort, but **`effort` is frontmatter-only and is NOT set at spawn** (platform limitation — effort cannot be overridden when spawning an agent), so override **only** `model`; effort stays whatever the agent's frontmatter declares.\n\nRe-run all gates after any fix until every one passes.\n\n---\n\n## Step 1b: Findings gate (load-bearing — blocks the ship)\n\nRuns **only after** the Step 1 quality gates all pass, and **before** Step 2 git readiness. The verification loop persists confirmed, blocking findings to a tracked `FINDINGS.md` (via `\u002Fverify`); this gate is the door where they become load-bearing. An open, blocking finding blocks `\u002Fship` exactly as a failing quality gate does — `FINDINGS.md` is enforcement, not advisory paperwork.\n\n**Resolve the brief dir (same resolution `\u002Fship` already uses for the PR body).** Find the active unit and mission, then glob the brief directory — the same dir that holds `FINDINGS.md`. Follow the repo's **one-resolver-per-helper** idiom (as `\u002Fverify` Step 6 does): resolve `$gm` against the helper each block actually calls. This block calls `godmode-state`, so it probes `godmode-state`; assert `$gm` is non-empty before any helper call — an empty resolver means the helper dir was not found, and the gate **fails closed** (BLOCK) rather than running an unresolved path:\n\n```bash\ngm=$(for c in \"${CLAUDE_PLUGIN_ROOT:-}\" \"$HOME\u002F.claude\" .; do [ -x \"$c\u002Fbin\u002Fgodmode-state\" ] && { echo \"$c\u002Fbin\"; break; }; done)\n[ -n \"$gm\" ] || { echo \"error: godmode helpers not found — BLOCK (fail-closed)\" >&2; exit 1; }\nactive_unit=$(\"$gm\u002Fgodmode-state\" get active_unit)\nmission_id=$(\"$gm\u002Fgodmode-state\" get mission_id)\nNN=$(printf '%02d' \"$active_unit\" 2>\u002Fdev\u002Fnull)\nbrief_dir=$(ls -d .planning\u002Fmissions\u002F${mission_id}\u002Fbriefs\u002F${NN}-* 2>\u002Fdev\u002Fnull | head -1)\n```\n\n**The skip-vs-block decision is a single shell predicate — never model judgment.** The gate's applicability and its block both turn on one deterministic shell decision that co-locates the explicit `[ -f \"$brief_dir\u002FFINDINGS.md\" ]` file test with the `count` read. Do **not** phrase the skip as a prose \"or it holds no `FINDINGS.md`\" model check — make it the shell predicate below.\n\n```bash\n# Resolve the findings helper against its own probe (one-resolver-per-helper).\ngm=$(for c in \"${CLAUDE_PLUGIN_ROOT:-}\" \"$HOME\u002F.claude\" .; do [ -x \"$c\u002Fbin\u002Fgodmode-findings\" ] && { echo \"$c\u002Fbin\"; break; }; done)\n[ -n \"$gm\" ] || { echo \"error: godmode helpers not found — BLOCK (fail-closed)\" >&2; exit 1; }\n\nif [ -z \"$brief_dir\" ]; then\n  # No active-unit brief dir → standalone change, nothing to gate on → SKIP.\n  : skip_findings_gate\nelif [ -f \"$brief_dir\u002FFINDINGS.md\" ]; then\n  # The store exists → read the live blocking count, the gate's sole truth.\n  count_status=0\n  open_blocking=$(\"$gm\u002Fgodmode-findings\" count \"$brief_dir\" --blocking) || count_status=$?\n  # Fail closed unless the count succeeded AND is a clean non-negative integer\n  # AND equals 0. A non-zero exit, a garbled\u002Fempty value (e.g. a resolver miss),\n  # or any value > 0 all collapse to one BLOCK predicate — never fail-open.\n  if [ \"$count_status\" -ne 0 ] || ! printf '%s' \"$open_blocking\" | grep -qE '^[0-9]+$' || [ \"$open_blocking\" -gt 0 ]; then\n    : BLOCK   # see the Decision below\n  else\n    : pass    # open_blocking == 0, count clean → gate passes\n  fi\nelse\n  # FINDINGS.md is ABSENT. Because .planning\u002F is gitignored\u002Flocal-only, an absent\n  # store is reachable in normal use (fresh clone, wiped local state) — and the\n  # frozen helper's `count --blocking` returns 0\u002Fexit-0 for an absent store, so a\n  # count read here would fail OPEN. Corroborate against the verify-recorded\n  # advisory tripwire instead: if verify recorded open_blocking > 0 but the store\n  # is now gone, it was LOST after verification found blockers → BLOCK.\n  gm=$(for c in \"${CLAUDE_PLUGIN_ROOT:-}\" \"$HOME\u002F.claude\" .; do [ -x \"$c\u002Fbin\u002Fgodmode-state\" ] && { echo \"$c\u002Fbin\"; break; }; done)\n  [ -n \"$gm\" ] || { echo \"error: godmode-state not found — BLOCK (fail-closed)\" >&2; exit 1; }\n  prior_blocking=$(\"$gm\u002Fgodmode-state\" get open_blocking 2>\u002Fdev\u002Fnull)\n  if printf '%s' \"$prior_blocking\" | grep -qE '^[0-9]+$' && [ \"$prior_blocking\" -gt 0 ]; then\n    : BLOCK   # store was lost after verify recorded blockers — fail closed\n  else\n    # open_blocking unset\u002F0 and no FINDINGS.md → a unit not yet verified → SKIP.\n    : skip_findings_gate\n  fi\nfi\n```\n\nThe blocking predicate (`status==open AND (sev==CRITICAL OR conf==HIGH)`) is the frozen helper's authoritative definition — never re-derived here. This gate trusts the **live** count it just read as its **pass-decision source of truth**. It does **not** read the verify-recorded `open_blocking` state key to decide a *pass*: that key is **advisory only** (it lets `\u002Fgodmode` surface the right next command), and a `\u002Fbuild N --fix` since the last `\u002Fverify` could have left it stale.\n\n**The `open_blocking` tripwire only ever *adds* a BLOCK — it never suppresses one — so the gate stays AC-9-consistent.** `open_blocking` is consulted in exactly one place: the **absent-`FINDINGS.md`** branch, purely as a fail-closed tripwire that can turn a would-be skip into a BLOCK. It is never used as the count that decides a pass; when a `FINDINGS.md` exists, the live `count` is the only source and `open_blocking` is not read at all. So this missing-store corroboration can only strengthen the gate, never weaken it.\n\n**Not applicable → skip (the legitimate cases).** The gate is skipped only when there is genuinely nothing to gate on: **no** active-unit `brief_dir` (a standalone change — consistent with `\u002Fship` already omitting the PR's **Brief** section), **or** a `brief_dir` whose recorded `open_blocking` is unset\u002F`0` and which has no `FINDINGS.md` (a unit not yet verified). Every other path reaches the BLOCK-or-pass decision below.\n\n**Fail-closed, collapsed into one predicate.** The store-exists branch fails **closed** unless the `count` exited 0 **and** its output is a clean `^[0-9]+$` integer **and** that integer is `0`. A non-zero exit (store unreadable\u002Fcorrupt), a non-numeric\u002Fempty `count` (a garbled value or resolver miss), or any value `> 0` all collapse into the single BLOCK predicate above — treat the gate as **BLOCKED** and report the error; the gate never ships past a store it cannot cleanly read. The absent-store branch fails closed via the `open_blocking` tripwire as described above.\n\n**Decision:**\n\n- **Gate passes** → only when the store exists, `count` succeeded, its output is a clean integer, and `open_blocking == 0` → proceed to Step 2. (Or the legitimate skip cases above.)\n- **BLOCK** → any of: `count` exited non-zero; `count` output is non-numeric\u002Fempty; live `open_blocking > 0`; or `FINDINGS.md` absent while the recorded `open_blocking > 0`. A BLOCK carries the same hard semantics as a failing quality gate: do **not** commit, do **not** `git push`, do **not** `gh pr create`, do **not** record `status=shipped`, and never `--no-verify` or otherwise bypass. **This BLOCK applies in both normal and `--no-push` mode** — `--no-push` does not exempt the findings gate any more than it exempts the quality gates (under `--no-push` the BLOCK means: do not even commit).\n\nOn BLOCK, report the blocking findings so the user can act:\n\n```bash\n\"$gm\u002Fgodmode-findings\" list \"$brief_dir\" --open --blocking --decode\n```\n\n**The two escapes.** When blocked, present exactly two ways forward:\n\n1. **Fix** — run `\u002Fbuild N --fix` to mechanically resolve the open findings, then re-`\u002Fverify` and re-`\u002Fship`.\n2. **Waive** — if a blocking finding is a genuine false positive, waive it (reason required).\n\n**The waive is conversational — `\u002Fship` never auto-waives.** `\u002Fship` does **not** waive anything on its own and never invents or supplies a reason. **Only** when the user explicitly names a specific finding ID **and** a reason does `\u002Fship` run the waive:\n\n```bash\n\"$gm\u002Fgodmode-findings\" waive \"$brief_dir\" \"\u003CID-the-user-named>\" \"\u003Creason-the-user-gave>\"\n```\n\nThe helper rejects an empty reason (non-zero exit) — so a waive without a user-supplied reason cannot succeed. After each waive, **re-read the live count in the same invocation** and proceed the moment it reaches 0:\n\n```bash\nopen_blocking=$(\"$gm\u002Fgodmode-findings\" count \"$brief_dir\" --blocking)\n```\n\nA multi-finding block clears by repeating this conversational waive per finding (or by `\u002Fbuild N --fix`) — there is no bulk-waive. Once the live count reaches 0, the gate passes and `\u002Fship` proceeds to Step 2. Until then, the ship stays blocked.\n\n---\n\n## Step 2: Git readiness\n\n- Commit or stash any uncommitted changes — the working tree must be clean before pushing.\n- Confirm the branch is current with its base.\n- Review the commit history: atomic commits, clear messages. If messy, suggest a rebase (ask first; never rewrite without consent).\n\n**Force-push guard.** **Never force-push to `main` or `master`** (nor any protected base) without an **explicit user request**. A normal `\u002Fship` uses a fast-forward `git push`; if a non-fast-forward push to a base branch is ever required, stop and ask the user to confirm in plain words before using `--force` \u002F `--force-with-lease`.\n\n---\n\n## Step 3: Push and open the PR\n\n**If `--no-push` is set:** stop here. Draft the PR title\u002Fbody (see below) and print it for the user, then report that the branch is gate-green and committed but **not** pushed — skip Step 4 (do not record `status=shipped`). Tell the user to re-run `\u002Fship` without `--no-push` when ready to publish.\n\nAfter confirmation (or immediately, in Auto Mode):\n\n```bash\nbranch=$(git branch --show-current)\n# Guard: never ship FROM a base branch. If you're on main\u002Fmaster, you'd push\n# the base itself — stop and ask the user to switch to a feature branch.\ncase \"$branch\" in\n  main|master) echo \"On base branch '$branch' — switch to a feature branch before \u002Fship.\" >&2; exit 1 ;;\nesac\ngit push -u origin \"$branch\"\n```\n\nDraft the PR body from the work unit. When a brief is present, **link it**: `.planning\u002Fmissions\u002F\u003Cmission_id>\u002Fbriefs\u002FNN-*\u002FBRIEF.md` describes the why + what + spec for the active work unit (find it via `\"$gm\u002Fgodmode-state\" get active_unit` and `mission_id=$(\"$gm\u002Fgodmode-state\" get mission_id)` (resolve `$gm` per the helper-resolution note above), then glob `.planning\u002Fmissions\u002F${mission_id}\u002Fbriefs\u002FNN-*\u002FBRIEF.md`). Summarize the work unit in the PR body and reference the brief path.\n\n```bash\ngh pr create --title \"\u003Cconcise, \u003C70 chars>\" --body \"$(cat \u003C\u003C'EOF'\n## Summary\n- What this work unit delivers and why (2-3 bullets)\n\n## Brief\n- .planning\u002Fmissions\u002F\u003Cmission_id>\u002Fbriefs\u002FNN-name\u002FBRIEF.md (the why + what + spec)\n\n## Changes\n- The concrete changes in this branch\n\n## Verification\n- Quality gates passed (config\u002Fquality-gates.txt)\n- How to verify\nEOF\n)\"\n```\n\nIf no brief exists (e.g. a standalone change), omit the **Brief** section and summarize the work directly.\n\nReturn the PR URL.\n\n---\n\n## Step 4: Record workflow state\n\nOn success, point the workflow forward via the single state source — `bin\u002Fgodmode-state`:\n\n```bash\ngm=$(for c in \"${CLAUDE_PLUGIN_ROOT:-}\" \"$HOME\u002F.claude\" .; do [ -x \"$c\u002Fbin\u002Fgodmode-state\" ] && { echo \"$c\u002Fbin\"; break; }; done)\n\"$gm\u002Fgodmode-state\" set status \"shipped\"\n\"$gm\u002Fgodmode-state\" set next_command \"\u002Fmission\"\n```\n\nThis lets `\u002Fgodmode` tell the user the work unit shipped and what to do next.\n\n---\n\n## Output\n\n\u003C!-- Output follows the in-skill output convention: godmode:output-convention — see rules\u002Fgodmode-output.md -->\n\nAfter shipping, report:\n\n- The gate results (all ✓).\n- The pushed branch and the PR URL.\n- The brief that was linked (if any).\n- The workflow state recorded and the next step:\n\n> \"Shipped. PR opened at [URL]. Run `\u002Fmission` to pick up the next work unit.\"\n\n---\n\n## Related\n\n- **\u002Fverify N** — preceding step: confirm the work unit meets its brief before shipping.\n- **\u002Fdebug** — when a quality gate fails.\n- **\u002Fgodmode** — reads the workflow state this skill records and tells the user the next command.\n\n**Spine:** `\u002Fmission` → `\u002Fbrief N` → `\u002Fplan N` → `\u002Fbuild N` → `\u002Fverify N` → `\u002Fship`. Ship gates, pushes, and opens the PR — the last step before merge.\n","# Ship\n\nTake a verified work unit to a pull request: run the canonical quality gates, then — only if every gate passes — push the branch and open a PR. This is the final step of the spine: `\u002Fmission` → `\u002Fbrief N` → `\u002Fplan N` → `\u002Fbuild N` → `\u002Fverify N` → **`\u002Fship`**.\n\n`\u002Fship` is side-effecting (it pushes and opens a PR), so it is **user-triggered only** (`disable-model-invocation: true`). It is never auto-invoked.\n\n---\n\n## Flags\n\n- **`--no-push`** (aliases: `--dry-run`, `--local`) — run every quality gate **and the findings gate (Step 1b)**, prepare git readiness (commit), then **stop before the push and the PR**. Draft and print the PR title\u002Fbody for review, but do **not** run `git push` or `gh pr create`, and do **not** record `status=shipped`. Use this to land a verified, committed, gate-green branch locally without publishing it. `--no-push` does **not** exempt the findings gate any more than it exempts the quality gates: open blocking findings BLOCK `--no-push` too (no commit), exactly as they BLOCK a normal ship.\n\nWhen `--no-push` is set the confirmation prompt is moot — there is nothing side-effecting to confirm — so skip it and run the gates and git readiness straight through.\n\n---\n\n## Confirm by default\n\n`\u002Fship` **confirms with the user before the side-effecting steps** (the push and the `gh pr create`). This confirm follows the shared convention in `rules\u002Fgodmode-recommend.md` (`godmode:recommend-convention`) — its confirm\u002Fproceed rendering: lead with a **Recommended: proceed** line carrying a visible one-line rationale, then let the user override.\n\nAfter the gates pass and the PR body is drafted, show the branch, target, and PR title\u002Fbody, then lead with the recommended-proceed confirm before pushing — for example:\n\n```\nRecommended: proceed — every quality gate and the findings gate are green, and\nthe PR body is drafted. Reply \"go\" to push and open the PR, or name what to change.\n```\n\nWait for explicit confirmation before pushing.\n\n**Exception — Auto Mode.** When `## Auto Mode Active` is present in context, skip the confirmation prompt: proceed straight to push and PR create on the default choices, and treat any user course-correction as normal input. The quality-gate block (Step 1) **and the findings gate (Step 1b)** are never skipped, in either mode — Auto Mode skips only the confirmation, never a BLOCK. (The findings gate's waive escape is inherently conversational, so it still requires an explicit user-named ID + reason even under Auto Mode; Auto Mode never auto-waives.)\n\n---\n\n**Resolving the godmode helpers.** The `godmode-*` helpers live in the plugin install dir — **not** the consumer repo you're working in — so a bare `bin\u002Fgodmode-*` path fails from another project's working directory. Every `bash` block below that calls one resolves its location into `$gm` first (plugin mode → `$CLAUDE_PLUGIN_ROOT\u002Fbin`, manual install → `~\u002F.claude\u002Fbin`, in-repo → `.\u002Fbin`) and calls `\"$gm\u002Fgodmode-\u003Cname>\"`. Keep the resolver line; never call a helper by a bare relative path.\n\n## Step 1: Quality gates (read from the canonical source)\n\nThe gates are defined in **`config\u002Fquality-gates.txt`** — one gate per line. **Read that file; do not hardcode the gate list here.** It is the single source of truth, so the gates stay in sync with the rest of the plugin.\n\nDon't re-derive the gate-running loop: invoke the bundled **`skills\u002Fship\u002Fscripts\u002Fgates.sh`**. It resolves `config\u002Fquality-gates.txt` across install modes (`${CLAUDE_PLUGIN_ROOT}\u002Fconfig` → `~\u002F.claude\u002Fconfig` → repo-relative), reads the gate list, prints one labelled status line per gate, and exits non-zero if any gate fails. You supply the per-gate verdicts via `GATES_RUNNER`: a command `gates.sh` invokes once per gate as `$GATES_RUNNER \"\u003Cgate text>\"` — exit 0 = pass, non-zero = fail. Wire `GATES_RUNNER` to your auto-detected per-gate dispatcher: for each gate text, run the project's matching command (typecheck, lint, test, build, secret scan; lint includes `shellcheck` clean for any `.sh` change) and exit with its status.\n\nResolve the script the same way across install modes, then run it:\n\n```bash\n# Locate the bundled gate runner — plugin mode, manual install, then repo-relative.\nGATES_SH=\"\"\nfor cand in \\\n  \"${CLAUDE_PLUGIN_ROOT:-}\u002Fskills\u002Fship\u002Fscripts\u002Fgates.sh\" \\\n  \"$HOME\u002F.claude\u002Fskills\u002Fship\u002Fscripts\u002Fgates.sh\" \\\n  \"skills\u002Fship\u002Fscripts\u002Fgates.sh\"; do\n  [ -n \"$cand\" ] && [ -f \"$cand\" ] && { GATES_SH=\"$cand\"; break; }\ndone\n[ -n \"$GATES_SH\" ] || { echo \"error: gates.sh not found\" >&2; exit 1; }\n\n# GATES_RUNNER is your auto-detected per-gate dispatcher: it receives one gate's\n# text as $1, runs the project command that gate maps to, and exits 0 (pass) or\n# non-zero (fail). gates.sh reads config\u002Fquality-gates.txt and calls it per gate.\n# GATES_RUNNER must resolve to a single executable — a script path or a command\n# name on PATH — NOT a shell function (invisible to the child process) and NOT a\n# space-separated string like \"bash runner.sh\" (invoked as one token, so it\n# would fail to find that file). Point it at an executable runner script.\nGATES_RUNNER=\"$run_one_gate\" bash \"$GATES_SH\"\n```\n\n`gates.sh` prints the report below; map each gate text to its detected command in your `GATES_RUNNER` dispatcher:\n\n```\nQuality gates (from config\u002Fquality-gates.txt):\n  [✓\u002F✗] Typecheck passes (zero errors)\n  [✓\u002F✗] Lint passes (zero errors; shellcheck clean for any .sh change)\n  [✓\u002F✗] All tests pass (existing + new)\n  [✓\u002F✗] No hardcoded secrets in the diff\n  [✓\u002F✗] No regressions in related functionality\n  [✓\u002F✗] Changes match the original requirements\n```\n\n**BLOCK on any failure.** If a gate fails: stop, report **which** gate failed and the error, and do **not** push or open a PR. Never `--no-verify`, never bypass a gate.\n\n- Test \u002F typecheck failures → use `\u002Fdebug`.\n- Lint → auto-fix where available, otherwise fix manually.\n- Multi-file fixes → spawn `@writer`.\n\n**Model profile.** Before spawning `@writer` (or any agent), resolve the active model profile from `${CLAUDE_PLUGIN_OPTION_MODEL_PROFILE:-balanced}`, then call the resolver `\"$gm\u002Fgodmode-model\" \u003Cagent>` (resolve `$gm` as the **Resolving the godmode helpers** note above shows) to obtain the model for that agent under the active profile. Pass that model to the Agent tool's `model` override at spawn time. The resolver also reports the agent's effort, but **`effort` is frontmatter-only and is NOT set at spawn** (platform limitation — effort cannot be overridden when spawning an agent), so override **only** `model`; effort stays whatever the agent's frontmatter declares.\n\nRe-run all gates after any fix until every one passes.\n\n---\n\n## Step 1b: Findings gate (load-bearing — blocks the ship)\n\nRuns **only after** the Step 1 quality gates all pass, and **before** Step 2 git readiness. The verification loop persists confirmed, blocking findings to a tracked `FINDINGS.md` (via `\u002Fverify`); this gate is the door where they become load-bearing. An open, blocking finding blocks `\u002Fship` exactly as a failing quality gate does — `FINDINGS.md` is enforcement, not advisory paperwork.\n\n**Resolve the brief dir (same resolution `\u002Fship` already uses for the PR body).** Find the active unit and mission, then glob the brief directory — the same dir that holds `FINDINGS.md`. Follow the repo's **one-resolver-per-helper** idiom (as `\u002Fverify` Step 6 does): resolve `$gm` against the helper each block actually calls. This block calls `godmode-state`, so it probes `godmode-state`; assert `$gm` is non-empty before any helper call — an empty resolver means the helper dir was not found, and the gate **fails closed** (BLOCK) rather than running an unresolved path:\n\n```bash\ngm=$(for c in \"${CLAUDE_PLUGIN_ROOT:-}\" \"$HOME\u002F.claude\" .; do [ -x \"$c\u002Fbin\u002Fgodmode-state\" ] && { echo \"$c\u002Fbin\"; break; }; done)\n[ -n \"$gm\" ] || { echo \"error: godmode helpers not found — BLOCK (fail-closed)\" >&2; exit 1; }\nactive_unit=$(\"$gm\u002Fgodmode-state\" get active_unit)\nmission_id=$(\"$gm\u002Fgodmode-state\" get mission_id)\nNN=$(printf '%02d' \"$active_unit\" 2>\u002Fdev\u002Fnull)\nbrief_dir=$(ls -d .planning\u002Fmissions\u002F${mission_id}\u002Fbriefs\u002F${NN}-* 2>\u002Fdev\u002Fnull | head -1)\n```\n\n**The skip-vs-block decision is a single shell predicate — never model judgment.** The gate's applicability and its block both turn on one deterministic shell decision that co-locates the explicit `[ -f \"$brief_dir\u002FFINDINGS.md\" ]` file test with the `count` read. Do **not** phrase the skip as a prose \"or it holds no `FINDINGS.md`\" model check — make it the shell predicate below.\n\n```bash\n# Resolve the findings helper against its own probe (one-resolver-per-helper).\ngm=$(for c in \"${CLAUDE_PLUGIN_ROOT:-}\" \"$HOME\u002F.claude\" .; do [ -x \"$c\u002Fbin\u002Fgodmode-findings\" ] && { echo \"$c\u002Fbin\"; break; }; done)\n[ -n \"$gm\" ] || { echo \"error: godmode helpers not found — BLOCK (fail-closed)\" >&2; exit 1; }\n\nif [ -z \"$brief_dir\" ]; then\n  # No active-unit brief dir → standalone change, nothing to gate on → SKIP.\n  : skip_findings_gate\nelif [ -f \"$brief_dir\u002FFINDINGS.md\" ]; then\n  # The store exists → read the live blocking count, the gate's sole truth.\n  count_status=0\n  open_blocking=$(\"$gm\u002Fgodmode-findings\" count \"$brief_dir\" --blocking) || count_status=$?\n  # Fail closed unless the count succeeded AND is a clean non-negative integer\n  # AND equals 0. A non-zero exit, a garbled\u002Fempty value (e.g. a resolver miss),\n  # or any value > 0 all collapse to one BLOCK predicate — never fail-open.\n  if [ \"$count_status\" -ne 0 ] || ! printf '%s' \"$open_blocking\" | grep -qE '^[0-9]+$' || [ \"$open_blocking\" -gt 0 ]; then\n    : BLOCK   # see the Decision below\n  else\n    : pass    # open_blocking == 0, count clean → gate passes\n  fi\nelse\n  # FINDINGS.md is ABSENT. Because .planning\u002F is gitignored\u002Flocal-only, an absent\n  # store is reachable in normal use (fresh clone, wiped local state) — and the\n  # frozen helper's `count --blocking` returns 0\u002Fexit-0 for an absent store, so a\n  # count read here would fail OPEN. Corroborate against the verify-recorded\n  # advisory tripwire instead: if verify recorded open_blocking > 0 but the store\n  # is now gone, it was LOST after verification found blockers → BLOCK.\n  gm=$(for c in \"${CLAUDE_PLUGIN_ROOT:-}\" \"$HOME\u002F.claude\" .; do [ -x \"$c\u002Fbin\u002Fgodmode-state\" ] && { echo \"$c\u002Fbin\"; break; }; done)\n  [ -n \"$gm\" ] || { echo \"error: godmode-state not found — BLOCK (fail-closed)\" >&2; exit 1; }\n  prior_blocking=$(\"$gm\u002Fgodmode-state\" get open_blocking 2>\u002Fdev\u002Fnull)\n  if printf '%s' \"$prior_blocking\" | grep -qE '^[0-9]+$' && [ \"$prior_blocking\" -gt 0 ]; then\n    : BLOCK   # store was lost after verify recorded blockers — fail closed\n  else\n    # open_blocking unset\u002F0 and no FINDINGS.md → a unit not yet verified → SKIP.\n    : skip_findings_gate\n  fi\nfi\n```\n\nThe blocking predicate (`status==open AND (sev==CRITICAL OR conf==HIGH)`) is the frozen helper's authoritative definition — never re-derived here. This gate trusts the **live** count it just read as its **pass-decision source of truth**. It does **not** read the verify-recorded `open_blocking` state key to decide a *pass*: that key is **advisory only** (it lets `\u002Fgodmode` surface the right next command), and a `\u002Fbuild N --fix` since the last `\u002Fverify` could have left it stale.\n\n**The `open_blocking` tripwire only ever *adds* a BLOCK — it never suppresses one — so the gate stays AC-9-consistent.** `open_blocking` is consulted in exactly one place: the **absent-`FINDINGS.md`** branch, purely as a fail-closed tripwire that can turn a would-be skip into a BLOCK. It is never used as the count that decides a pass; when a `FINDINGS.md` exists, the live `count` is the only source and `open_blocking` is not read at all. So this missing-store corroboration can only strengthen the gate, never weaken it.\n\n**Not applicable → skip (the legitimate cases).** The gate is skipped only when there is genuinely nothing to gate on: **no** active-unit `brief_dir` (a standalone change — consistent with `\u002Fship` already omitting the PR's **Brief** section), **or** a `brief_dir` whose recorded `open_blocking` is unset\u002F`0` and which has no `FINDINGS.md` (a unit not yet verified). Every other path reaches the BLOCK-or-pass decision below.\n\n**Fail-closed, collapsed into one predicate.** The store-exists branch fails **closed** unless the `count` exited 0 **and** its output is a clean `^[0-9]+$` integer **and** that integer is `0`. A non-zero exit (store unreadable\u002Fcorrupt), a non-numeric\u002Fempty `count` (a garbled value or resolver miss), or any value `> 0` all collapse into the single BLOCK predicate above — treat the gate as **BLOCKED** and report the error; the gate never ships past a store it cannot cleanly read. The absent-store branch fails closed via the `open_blocking` tripwire as described above.\n\n**Decision:**\n\n- **Gate passes** → only when the store exists, `count` succeeded, its output is a clean integer, and `open_blocking == 0` → proceed to Step 2. (Or the legitimate skip cases above.)\n- **BLOCK** → any of: `count` exited non-zero; `count` output is non-numeric\u002Fempty; live `open_blocking > 0`; or `FINDINGS.md` absent while the recorded `open_blocking > 0`. A BLOCK carries the same hard semantics as a failing quality gate: do **not** commit, do **not** `git push`, do **not** `gh pr create`, do **not** record `status=shipped`, and never `--no-verify` or otherwise bypass. **This BLOCK applies in both normal and `--no-push` mode** — `--no-push` does not exempt the findings gate any more than it exempts the quality gates (under `--no-push` the BLOCK means: do not even commit).\n\nOn BLOCK, report the blocking findings so the user can act:\n\n```bash\n\"$gm\u002Fgodmode-findings\" list \"$brief_dir\" --open --blocking --decode\n```\n\n**The two escapes.** When blocked, present exactly two ways forward:\n\n1. **Fix** — run `\u002Fbuild N --fix` to mechanically resolve the open findings, then re-`\u002Fverify` and re-`\u002Fship`.\n2. **Waive** — if a blocking finding is a genuine false positive, waive it (reason required).\n\n**The waive is conversational — `\u002Fship` never auto-waives.** `\u002Fship` does **not** waive anything on its own and never invents or supplies a reason. **Only** when the user explicitly names a specific finding ID **and** a reason does `\u002Fship` run the waive:\n\n```bash\n\"$gm\u002Fgodmode-findings\" waive \"$brief_dir\" \"\u003CID-the-user-named>\" \"\u003Creason-the-user-gave>\"\n```\n\nThe helper rejects an empty reason (non-zero exit) — so a waive without a user-supplied reason cannot succeed. After each waive, **re-read the live count in the same invocation** and proceed the moment it reaches 0:\n\n```bash\nopen_blocking=$(\"$gm\u002Fgodmode-findings\" count \"$brief_dir\" --blocking)\n```\n\nA multi-finding block clears by repeating this conversational waive per finding (or by `\u002Fbuild N --fix`) — there is no bulk-waive. Once the live count reaches 0, the gate passes and `\u002Fship` proceeds to Step 2. Until then, the ship stays blocked.\n\n---\n\n## Step 2: Git readiness\n\n- Commit or stash any uncommitted changes — the working tree must be clean before pushing.\n- Confirm the branch is current with its base.\n- Review the commit history: atomic commits, clear messages. If messy, suggest a rebase (ask first; never rewrite without consent).\n\n**Force-push guard.** **Never force-push to `main` or `master`** (nor any protected base) without an **explicit user request**. A normal `\u002Fship` uses a fast-forward `git push`; if a non-fast-forward push to a base branch is ever required, stop and ask the user to confirm in plain words before using `--force` \u002F `--force-with-lease`.\n\n---\n\n## Step 3: Push and open the PR\n\n**If `--no-push` is set:** stop here. Draft the PR title\u002Fbody (see below) and print it for the user, then report that the branch is gate-green and committed but **not** pushed — skip Step 4 (do not record `status=shipped`). Tell the user to re-run `\u002Fship` without `--no-push` when ready to publish.\n\nAfter confirmation (or immediately, in Auto Mode):\n\n```bash\nbranch=$(git branch --show-current)\n# Guard: never ship FROM a base branch. If you're on main\u002Fmaster, you'd push\n# the base itself — stop and ask the user to switch to a feature branch.\ncase \"$branch\" in\n  main|master) echo \"On base branch '$branch' — switch to a feature branch before \u002Fship.\" >&2; exit 1 ;;\nesac\ngit push -u origin \"$branch\"\n```\n\nDraft the PR body from the work unit. When a brief is present, **link it**: `.planning\u002Fmissions\u002F\u003Cmission_id>\u002Fbriefs\u002FNN-*\u002FBRIEF.md` describes the why + what + spec for the active work unit (find it via `\"$gm\u002Fgodmode-state\" get active_unit` and `mission_id=$(\"$gm\u002Fgodmode-state\" get mission_id)` (resolve `$gm` per the helper-resolution note above), then glob `.planning\u002Fmissions\u002F${mission_id}\u002Fbriefs\u002FNN-*\u002FBRIEF.md`). Summarize the work unit in the PR body and reference the brief path.\n\n```bash\ngh pr create --title \"\u003Cconcise, \u003C70 chars>\" --body \"$(cat \u003C\u003C'EOF'\n## Summary\n- What this work unit delivers and why (2-3 bullets)\n\n## Brief\n- .planning\u002Fmissions\u002F\u003Cmission_id>\u002Fbriefs\u002FNN-name\u002FBRIEF.md (the why + what + spec)\n\n## Changes\n- The concrete changes in this branch\n\n## Verification\n- Quality gates passed (config\u002Fquality-gates.txt)\n- How to verify\nEOF\n)\"\n```\n\nIf no brief exists (e.g. a standalone change), omit the **Brief** section and summarize the work directly.\n\nReturn the PR URL.\n\n---\n\n## Step 4: Record workflow state\n\nOn success, point the workflow forward via the single state source — `bin\u002Fgodmode-state`:\n\n```bash\ngm=$(for c in \"${CLAUDE_PLUGIN_ROOT:-}\" \"$HOME\u002F.claude\" .; do [ -x \"$c\u002Fbin\u002Fgodmode-state\" ] && { echo \"$c\u002Fbin\"; break; }; done)\n\"$gm\u002Fgodmode-state\" set status \"shipped\"\n\"$gm\u002Fgodmode-state\" set next_command \"\u002Fmission\"\n```\n\nThis lets `\u002Fgodmode` tell the user the work unit shipped and what to do next.\n\n---\n\n## Output\n\n\u003C!-- Output follows the in-skill output convention: godmode:output-convention — see rules\u002Fgodmode-output.md -->\n\nAfter shipping, report:\n\n- The gate results (all ✓).\n- The pushed branch and the PR URL.\n- The brief that was linked (if any).\n- The workflow state recorded and the next step:\n\n> \"Shipped. PR opened at [URL]. Run `\u002Fmission` to pick up the next work unit.\"\n\n---\n\n## Related\n\n- **\u002Fverify N** — preceding step: confirm the work unit meets its brief before shipping.\n- **\u002Fdebug** — when a quality gate fails.\n- **\u002Fgodmode** — reads the workflow state this skill records and tells the user the next command.\n\n**Spine:** `\u002Fmission` → `\u002Fbrief N` → `\u002Fplan N` → `\u002Fbuild N` → `\u002Fverify N` → `\u002Fship`. Ship gates, pushes, and opens the PR — the last step before merge.\n",{"name":5,"description":7,"allowed-tools":43,"user-invocable":44,"disable-model-invocation":44},"Read, Glob, Grep, Bash(git *), Bash(gh *), Bash(gm=*), Bash(*godmode-state*), Bash(*godmode-model*), Bash(*godmode-findings*), Bash(*\u002Fskills\u002Fship\u002Fscripts\u002Fgates.sh*), Bash(skills\u002Fship\u002Fscripts\u002Fgates.sh*)",true,"main","717a1f3af996804502df0047840b6c74a1d464c5","skills\u002Fship",0,"https:\u002F\u002Fgithub.com\u002FSyloRei\u002Fclaude-godmode","https:\u002F\u002Fgithub.com\u002FSyloRei","Nya? Kawai? Desu?","2026-06-08 16:03:07",{"source":8,"skillName":5,"commands":54,"prompt":64,"zipAvailable":44},[55,58,61],{"tool":56,"command":57},"npx","npx skills add SyloRei\u002Fclaude-godmode --skill ship",{"tool":59,"command":60},"bunx","bunx skills add SyloRei\u002Fclaude-godmode --skill ship",{"tool":62,"command":63},"pnpm","pnpm dlx skills add SyloRei\u002Fclaude-godmode --skill ship","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FSyloRei\u002Fclaude-godmode\u002Fblob\u002Fmain\u002Fskills\u002Fship\u002FSKILL.md",[66],{"path":67,"resourceType":68,"resourceTypeDesc":69,"sizeBytes":70,"sourceHtmlUrl":71},"scripts\u002Fgates.sh","SCRIPT","脚本",4756,"https:\u002F\u002Fgithub.com\u002FSyloRei\u002Fclaude-godmode\u002Fblob\u002Fmain\u002Fskills\u002Fship\u002Fscripts\u002Fgates.sh",[73,82,101,123,152,177],{"skillNo":74,"name":75,"nameZh":6,"description":76,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":77,"categories":78,"keywords":79,"publishedAt":38},"01M6HBW3WEVBXANCRWF5YK5BYC","{{name=null}=null}","{{description=null}=null}","https:\u002F\u002Fgithub.com\u002FSyloRei\u002Fclaude-godmode\u002Fblob\u002Fmain\u002Ftemplates\u002Fskill\u002FSKILL.md",[],[75,80,81],"name","true",{"skillNo":83,"name":84,"nameZh":6,"description":85,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":86,"categories":87,"keywords":88,"publishedAt":38},"01M6HBVZ86GN3TGGFSJ4XM9C5Q","verify","Check a built unit goal-backward: classify every brief acceptance criterion COVERED \u002F PARTIAL \u002F MISSING with concrete evidence, so you know what is truly done. Use when: verify N, check whether unit N is done, confirm the goals for roadmap unit N are met.","https:\u002F\u002Fgithub.com\u002FSyloRei\u002Fclaude-godmode\u002Fblob\u002Fmain\u002Fskills\u002Fverify\u002FSKILL.md",[],[84,35,33,23,89,90,91,21,92,93,94,95,96,97,98,25,28,19,99,100],"you","ShellCheck","unit","classify","done","show","log","goal","roadmap","coverage","brief","acceptance",{"skillNo":102,"name":103,"nameZh":6,"description":104,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":105,"categories":106,"keywords":107,"publishedAt":38},"01M6HBVTDPNPD88ZWXT4T5M8EA","triage","Respond to a production incident, outage, or regression in the wild: establish the timeline, assess blast radius, and reconstruct what happened with cited evidence. Use this when something is failing in production and you need to understand it before fixing it.","https:\u002F\u002Fgithub.com\u002FSyloRei\u002Fclaude-godmode\u002Fblob\u002Fmain\u002Fskills\u002Ftriage\u002FSKILL.md",[],[103,108,109,81,89,110,111,112,113,114,115,116,117,118,119,120,121,122],"incident","wild","with","understand","it","respond","assess","production incident","evidence","use","regression","radius","timeline","Production","blast radius",{"skillNo":124,"name":125,"nameZh":6,"description":126,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":127,"categories":128,"keywords":137,"publishedAt":38},"01M6HBVNQ1ASFTJTYT47F8M559","tdd","Build new behavior test-first using strict Red-Green-Refactor cycles. Use this when you want to drive a new feature or behavior from failing tests before writing the implementation.","https:\u002F\u002Fgithub.com\u002FSyloRei\u002Fclaude-godmode\u002Fblob\u002Fmain\u002Fskills\u002Ftdd\u002FSKILL.md",[129,134],{"categoryNo":130,"categoryName":131,"taxonomyType":132,"taxonomyTypeDesc":133},"01JBSEEDCAT000000000000007","代码审查","SCENARIO","使用场景",{"categoryNo":135,"categoryName":136,"taxonomyType":132,"taxonomyTypeDesc":133},"01JBSEEDCAT000000000000023","Skill与智能体开发",[125,81,89,138,139,140,141,142,143,144,145,146,117,147,148,149,150,151],"red","first","feature","tests","green","build","implementation","drive","red-green-refactor","new feature","behavior","refactor","test","writing",{"skillNo":153,"name":154,"nameZh":6,"description":155,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":156,"categories":157,"keywords":163,"publishedAt":38},"01M6HBVB6SVSMQPXBMWEWQ4J5M","refine","Mid-mission gap analysis: read the active mission's roadmap and briefs, surface what's missing or underspecified, converge on one gap, and append a NEW numbered work unit plus its full brief — strictly additive. Use when: refine, what are we missing, find gaps in the current mission, add the next unit. Never edits an existing brief or roadmap unit in place — that is \u002Fbrief N's job.","https:\u002F\u002Fgithub.com\u002FSyloRei\u002Fclaude-godmode\u002Fblob\u002Fmain\u002Fskills\u002Frefine\u002FSKILL.md",[158],{"categoryNo":159,"categoryName":160,"taxonomyType":161,"taxonomyTypeDesc":162},"01JBSEEDCAT000000000000031","产品经理","PROFESSION","适用职业",[154,164,165,166,23,97,167,168,169,170,25,171,19,172,99,173,174,175,91,176],"work","active","next","never","append","edit","write","full","analysis","mission","converge","briefs","add",{"skillNo":178,"name":149,"nameZh":6,"description":179,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":180,"categories":181,"keywords":187,"publishedAt":38},"01M6HBV7SGBH14T5K2336RYZWW","Improve code structure without changing behavior, verifying with tests at every step. Use this when you want to restructure, extract, rename, or simplify existing code that already works.","https:\u002F\u002Fgithub.com\u002FSyloRei\u002Fclaude-godmode\u002Fblob\u002Fmain\u002Fskills\u002Frefactor\u002FSKILL.md",[182,183,184],{"categoryNo":130,"categoryName":131,"taxonomyType":132,"taxonomyTypeDesc":133},{"categoryNo":135,"categoryName":136,"taxonomyType":132,"taxonomyTypeDesc":133},{"categoryNo":185,"categoryName":186,"taxonomyType":161,"taxonomyTypeDesc":162},"01JBSEEDCAT000000000000030","开发工程师",[149,148,188,189,190,117,191,192,81,89,110,141,193,194,195],"code","simplify","improve","Existing","at","structure","rename","extract"]