[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7AB2Z0JEY4BEC05ZWAYFT0E":3,"skill-same:01M7AB2Z0JEY4BEC05ZWAYFT0E":43},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":15,"categories":16,"keywords":17,"publishedAt":18,"linkOnly":19,"markdownContent":20,"markdownBody":21,"frontMatter":22,"refName":23,"commitSha":24,"skillDirectory":25,"forksCount":13,"repositoryHtmlUrl":26,"repositoryHomepageUrl":6,"ownerHtmlUrl":27,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":28,"install":29,"resources":42},"01M7AB2Z0JEY4BEC05ZWAYFT0E","ha-issue-triage",null,"Work through a backlog of GitHub issues on this repository — categorize, then walk bugs and feature requests one at a time, fixing approved ones on a branch through a pull request instead of committing straight to main. Use when asked to \"triage my issues\", \"go through the issue backlog\", \"work through the issues\", \"let's triage\", or invoked via the `\u002Fissues` command. This is opt-in tooling for backlog work, not a replacement for the normal ad-hoc flow in AGENTS.md — a single \"fix this bug\" request in conversation still goes through that flow, not this one. SYMPTOMS — load this if you are about to: fix several issues in a row by committing each straight to main with nobody reviewing between commits; dump the whole backlog in one message instead of walking it one issue at a time; merge a pull request this flow opened without either explicit confirmation or an explicit full-autonomy invocation; bump a version by hand; or skip the breaking-changes warn-first check on an issue fix that needs it.","amg0\u002Fha_ksenia","amg0","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F16444411?v=4","01M7AAZDRNF8BXHM6GAN9QWJ41","01M7AAZDSEHKPEWK6PKK22S2NN",0,"MIT","https:\u002F\u002Fgithub.com\u002Famg0\u002Fha_ksenia\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fha-issue-triage\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: ha-issue-triage\ndescription: >-\n  Work through a backlog of GitHub issues on this repository — categorize, then walk bugs and feature requests one\n  at a time, fixing approved ones on a branch through a pull request instead of committing straight to main. Use\n  when asked to \"triage my issues\", \"go through the issue backlog\", \"work through the issues\", \"let's triage\",\n  or invoked via the `\u002Fissues` command. This is opt-in tooling for backlog work, not a replacement for the normal\n  ad-hoc flow in AGENTS.md — a single \"fix this bug\" request in conversation still goes through that flow, not this\n  one. SYMPTOMS — load this if you are about to: fix several issues in a row by committing each straight to main\n  with nobody reviewing between commits; dump the whole backlog in one message instead of walking it one issue at a\n  time; merge a pull request this flow opened without either explicit confirmation or an explicit full-autonomy\n  invocation; bump a version by hand; or skip the breaking-changes warn-first check on an issue fix that needs it.\n---\n\n# Issue triage and branch\u002FPR-based fixing\n\nA backlog-scale companion to the ad-hoc flow in `AGENTS.md`. That flow (commit directly, on explicit request, no\npush) is still correct for a single request handled live with the developer watching. This skill is for the\ndifferent situation: working through several issues in a row, where a branch + pull request + green CI is the\nreview surface, because nobody is watching each individual commit in real time. Both are legitimate; pick based on\nwhether you're handling one request or a backlog, not out of habit.\n\n## Merge autonomy — read this before the fix flow below\n\nBy default, **ask for explicit confirmation before merging** any pull request this flow opens — merging is a\nshared-state, hard-to-reverse action (`AGENTS.md`, \"Executing actions with care\"), and a backlog run touches several\nof them in a row.\n\nSkip that per-PR confirmation, and merge automatically on green CI, only when this session was entered through a\ndedicated entry point built for full autonomy — a command that exists specifically to invoke this flow (such as\n`\u002Fissues`, if your client offers slash commands), or an explicit instruction scoped to the session (\"auto-merge\nthese\", \"don't ask me each time today\"). Ad-hoc conversational use of this skill — \"hey, can you triage my issues\" —\nalways asks. State which mode you're in once, at the start of a run, rather than leaving it implicit.\n\n## Concurrency: running more than one instance on the same repo\n\nIf nothing else is touching this backlog concurrently, skip this section — claim\u002Frelease below are harmless no-ops.\nIt exists for when you might be one of several agent sessions working the same repo at once.\n\n**Instance ID:** generate one ID of the form `\u003Cepoch>-\u003Crand>` (`` `date +%s` `` seconds, plus a random suffix) at the\nstart of a run, show it, and reuse the literal value in every later command — shell variables do not survive\nbetween separate tool calls. The lock label is `wip:\u003Cid>`; the epoch doubles as its creation timestamp for the TTL\nbelow.\n\n**Locked** means an issue carries a _foreign_ `wip:\u003Cepoch>-*` label whose epoch is younger than 6 hours. Don't touch\nthose — list them, move on. A foreign lock older than 6 hours is orphaned and does not lock (see \"Expired locks\").\n\n**Claim** (before any write to an issue — categorize, fix, comment, snooze; plain viewing needs no lock):\n\n1. `gh issue view N --json labels` — a foreign non-expired `wip:*` already there? It's taken; say so, move on.\n2. Ensure both labels exist with `gh label create \"in progress\" --color fbca04 --force` and\n   `gh label create \"wip:\u003Cid>\" --color 5319e7 --force`, then\n   `gh issue edit N --add-label \"wip:\u003Cid>\" --add-label \"in progress\"`.\n3. **Race check:** view the labels again. Several `wip:*` present → the lexicographically smallest ID wins. Not the\n   smallest → remove your label, tell the developer, move on. Otherwise you hold the lock.\n\n**Release** immediately after a quick write (categorize\u002Fcomment\u002Fsnooze); hold it through a fix until merge. Release\nwith `gh issue edit N --remove-label \"wip:\u003Cid>\" --remove-label \"in progress\"` (a merged fix closes the issue anyway).\n\n**Expired locks (TTL 6h):** `now − epoch > 21600` → orphaned. Remove the foreign label, then claim normally with your\nown ID.\n\n**Label cleanup:** whenever a `wip:\u003Cid>` label is removed (release, race back-off, `unlock`, expired takeover) and no\nopen issue still carries it, delete the label definition too (`gh label delete \"wip:\u003Cid>\" --yes`) so `wip:*` doesn't\npile up.\n\n## Dispatch\n\n- `locks` — list open issues carrying a `wip:*` label (number, title, holding ID); mark ones older than 6h\n  \"(expired, treated as free)\". Nothing else.\n- `unlock \u003Cissue#>` — show the issue's `wip:*`\u002F`in progress` labels, confirm, remove them.\n- a bare issue number — skip triage, show the issue and its comments, get the developer's read on it before writing\n  any code (plan mode \u002F a quick back-and-forth, not straight into the fix flow).\n- `stop` \u002F `drain` (as an argument, or said mid-run) — see \"Graceful stop\" below.\n- no argument — the full flow below (generate the instance ID first if concurrency applies).\n\n## Phase 0 — prerequisites and scope (counts only)\n\nList the repository labels first. Create the workflow label once with\n`gh label create \"in progress\" --color fbca04 --force`; claims depend on it and must not fail halfway through. Resolve\nthe two category labels from what is actually present: prefer `bug` for defects and `enhancement` for feature work,\nfall back to `Feature request` if that is the repository's convention, and ask when several plausible labels remain.\nReuse those exact names for the whole run.\n\nThen run `gh issue list --state open --limit 1000 --json number,title,labels,assignees,author`. Run the snooze wake pass\n(see below) and set aside issues with a foreign, non-expired `wip:*`. Report only counts — how many need categorizing,\nhow many bugs, how many feature requests, one line each for anything that woke up or is locked elsewhere. Do not\nenumerate the backlog; go straight into the per-issue walk. If the repository has more than 1,000 open issues, say\nthat this is the first batch rather than presenting the counts as complete.\n\n## Phase 1 — categorize\n\nUse the defect and feature labels resolved in Phase 0. Walk the uncategorized issues one at a time:\n\n1. A short summary (2–3 sentences) of what the issue wants.\n2. A weak title? Suggest a better one.\n3. Ask about **only this issue**, with a reasoned recommendation (bug vs. feature request; `question`\u002F`duplicate`\u002F\n   `wontfix` when in doubt), fold an optional title change into the same question. Wait for the answer.\n4. After the decision: claim, `gh issue edit N --add-label \"\u003Clabel>\"` (and `--title` if changed), release.\n\n## Phase 2 \u002F 3 — walk the bugs, then the feature requests\n\nOne at a time, waiting for the developer after each:\n\n1. `gh issue view N --comments` — explain the cause\u002Fcontext and give an honest read (worth fixing, effort, risk,\n   possible duplicate).\n2. If there are comments, summarize the thread — who said what, what's open. No comments → skip this step entirely.\n3. Ask what should happen, with a recommendation: **Fix** (below) \u002F **Ask the author** \u002F **Reject \u002F decline** \u002F\n   **Snooze** \u002F **Skip** (next issue, release any held lock).\n\nFor a feature request, \"Fix\" means \"implement,\" and scope is more often unsettled — clarify with the author or with\n`ha-planning` before code appears. Name dependencies between related requests when you see them.\n\n## Fix flow (issue approved)\n\n1. **Claim.**\n2. **Breaking-changes check first.** If the fix touches anything on `AGENTS.md`'s breaking-changes list (unique\n   IDs, entity IDs, entry data, state values, service signatures, removing a config option), stop and follow\n   [`ha-breaking-changes`](..\u002Fha-breaking-changes\u002FSKILL.md)'s warn-first procedure before writing any code — vutuv,\n   the source this skill generalizes from, has no equivalent concern, but this project does.\n3. **Isolate and branch.** Run `git status --short` before creating anything. If the current worktree is dirty, do\n   not absorb, stash, or discard those changes: create a separate worktree from `main`, or stop if an isolated\n   worktree is not available. In a clean tree, create a short kebab-case branch; never commit the fix on `main` in\n   this flow.\n4. **Implement.** If your agent runtime supports isolated sub-agent dispatch, use it; otherwise work directly on the\n   branch. For a bug, write the test that reproduces it first and watch it fail, then fix until it passes — a test\n   written after the fix may never have caught the bug. Other tests stay proportionate, per\n   [`ha-testing`](..\u002Fha-testing\u002FSKILL.md). If the fix turns out to need more than ~10 files or an architectural\n   change, stop and hand off to [`ha-planning`](..\u002Fha-planning\u002FSKILL.md) instead of continuing inline.\n5. **Gate once.** Run `script\u002Fcheck` (type-check + lint-check + spell-check) redirected to a log, judged by its exit\n   code — see [`blueprint-tooling`](..\u002Fblueprint-tooling\u002FSKILL.md)'s token-discipline note. Read the log, and only\n   the failing part, on a non-zero exit; iterate on just the failing piece, then re-run the full gate once more\n   before moving on.\n6. **Commit** using Conventional Commits\n   ([`blueprint.commit-message.instructions.md`](..\u002F..\u002Finstructions\u002Fblueprint.commit-message.instructions.md)).\n   **No manual version bump** — release-please derives it from the commit type once this lands on `main`\n   ([`ha-release`](..\u002Fha-release\u002FSKILL.md)).\n7. **Push and open the PR** (`gh pr create`, following `.github\u002Fpull_request_template.md` if the repo has one).\n   Open it ready for review, not as a draft: the gate above already passed, and CI skips drafts, so a draft would\n   leave the next step waiting for checks that never start.\n8. **Wait for CI:** `gh pr checks --watch --fail-fast`.\n   - Red → do not merge. Fix it and push again, watch again — or stop and report if it isn't yours to fix.\n   - Green → continue.\n9. **Merge**, per the autonomy rule at the top of this file. `gh pr merge \u003Cnr> --squash --delete-branch`.\n10. **Branch cleanup — every time, not optional.** `gh pr merge --delete-branch` normally removes both copies of the\n    branch. Refresh `main`, prune, and delete the local branch only if it still exists:\n\n    ```bash\n    git checkout main && git pull --ff-only\n    git fetch --prune origin\n    git show-ref --verify --quiet refs\u002Fheads\u002F\u003Cbranch> && git branch -D \u003Cbranch>\n    ```\n\n    The final command returning 1 because the branch is already gone is success, not a cleanup failure. If the fix\n    ran in an isolated worktree, remove that worktree before checking for the branch. Before moving on,\n    `git branch -vv` should show nothing marked `[origin\u002F\u003Cname>: gone]` — that marker is the leak.\n\n11. **Release the claim.** Remove `wip:\u003Cid>` and `in progress` even if the merge already closed the issue, then delete\n    the now-unused `wip:\u003Cid>` label definition.\n12. **Close the loop on the issue.** The merge may auto-close it (a `Fixes #N` PR body); either way, draft a short\n    shipped note — what now works, and why, in the language of the issue — per `AGENTS.md`'s \"Posting on the\n    developer's behalf\": get explicit sign-off on the draft unless this session is running in full-autonomy mode,\n    and disclose per `AI_POLICY.md` when it posts under the developer's identity without another marker.\n\n## Ask the author\n\nAlways an option in the per-issue walk. Get the author (`gh issue view N --json author`), draft a friendly question\nin their language, following the same posting-on-behalf policy as step 11 above. Claim → comment → release.\n\n## Reject \u002F decline\n\nWhen the developer decides an issue won't be done — out of scope, a design not wanted, a duplicate — close it\nkindly, never leave it to rot. Ask for the reason as part of the same question (a short recommendation is fine; \"no\nreason, just decline\" is a valid answer), then, after claiming: draft the closing note per the posting-on-behalf\npolicy, get sign-off, comment, `gh issue close N --reason \"not planned\"` (`--reason completed` only when it's\ngenuinely resolved another way), and release — a closed issue keeps no lock.\n\n## Snooze\n\nLabels `snoozed` + `snooze:YYYY-MM-DD` (name = wake date). After claiming: `gh label create snoozed --color c5def5\n--force`, `gh label create \"snooze:\u003Cdate>\" --color ededed --force`, apply both, release. At the start of every run,\ndo a wake pass: `snooze:\u003Cdate>` ≤ today → remove both labels, surface it at the top as woken up; delete now-empty\n`snooze:*` labels afterward. Bundle future ones into one line (\"N snoozed\").\n\n## Graceful stop (drain)\n\nThe developer can end a run at any point — not an emergency kill. **Finish work already in flight, start nothing\nnew:**\n\n1. Stop claiming, categorizing, commenting, snoozing, or dispatching anything new. The per-issue walk ends here.\n2. Let an in-flight fix finish: when it reports, apply the normal merge rule (green → merge + cleanup + release;\n   red\u002Funclear → hold the lock and report).\n3. For anything claimed but not yet started, list it and ask once whether to keep the claim for next time or release\n   it now (default recommendation: release the lock, keep any assignee).\n4. Report what shipped this run, what's still draining, and what was left untouched.\n","# Issue triage and branch\u002FPR-based fixing\n\nA backlog-scale companion to the ad-hoc flow in `AGENTS.md`. That flow (commit directly, on explicit request, no\npush) is still correct for a single request handled live with the developer watching. This skill is for the\ndifferent situation: working through several issues in a row, where a branch + pull request + green CI is the\nreview surface, because nobody is watching each individual commit in real time. Both are legitimate; pick based on\nwhether you're handling one request or a backlog, not out of habit.\n\n## Merge autonomy — read this before the fix flow below\n\nBy default, **ask for explicit confirmation before merging** any pull request this flow opens — merging is a\nshared-state, hard-to-reverse action (`AGENTS.md`, \"Executing actions with care\"), and a backlog run touches several\nof them in a row.\n\nSkip that per-PR confirmation, and merge automatically on green CI, only when this session was entered through a\ndedicated entry point built for full autonomy — a command that exists specifically to invoke this flow (such as\n`\u002Fissues`, if your client offers slash commands), or an explicit instruction scoped to the session (\"auto-merge\nthese\", \"don't ask me each time today\"). Ad-hoc conversational use of this skill — \"hey, can you triage my issues\" —\nalways asks. State which mode you're in once, at the start of a run, rather than leaving it implicit.\n\n## Concurrency: running more than one instance on the same repo\n\nIf nothing else is touching this backlog concurrently, skip this section — claim\u002Frelease below are harmless no-ops.\nIt exists for when you might be one of several agent sessions working the same repo at once.\n\n**Instance ID:** generate one ID of the form `\u003Cepoch>-\u003Crand>` (`` `date +%s` `` seconds, plus a random suffix) at the\nstart of a run, show it, and reuse the literal value in every later command — shell variables do not survive\nbetween separate tool calls. The lock label is `wip:\u003Cid>`; the epoch doubles as its creation timestamp for the TTL\nbelow.\n\n**Locked** means an issue carries a _foreign_ `wip:\u003Cepoch>-*` label whose epoch is younger than 6 hours. Don't touch\nthose — list them, move on. A foreign lock older than 6 hours is orphaned and does not lock (see \"Expired locks\").\n\n**Claim** (before any write to an issue — categorize, fix, comment, snooze; plain viewing needs no lock):\n\n1. `gh issue view N --json labels` — a foreign non-expired `wip:*` already there? It's taken; say so, move on.\n2. Ensure both labels exist with `gh label create \"in progress\" --color fbca04 --force` and\n   `gh label create \"wip:\u003Cid>\" --color 5319e7 --force`, then\n   `gh issue edit N --add-label \"wip:\u003Cid>\" --add-label \"in progress\"`.\n3. **Race check:** view the labels again. Several `wip:*` present → the lexicographically smallest ID wins. Not the\n   smallest → remove your label, tell the developer, move on. Otherwise you hold the lock.\n\n**Release** immediately after a quick write (categorize\u002Fcomment\u002Fsnooze); hold it through a fix until merge. Release\nwith `gh issue edit N --remove-label \"wip:\u003Cid>\" --remove-label \"in progress\"` (a merged fix closes the issue anyway).\n\n**Expired locks (TTL 6h):** `now − epoch > 21600` → orphaned. Remove the foreign label, then claim normally with your\nown ID.\n\n**Label cleanup:** whenever a `wip:\u003Cid>` label is removed (release, race back-off, `unlock`, expired takeover) and no\nopen issue still carries it, delete the label definition too (`gh label delete \"wip:\u003Cid>\" --yes`) so `wip:*` doesn't\npile up.\n\n## Dispatch\n\n- `locks` — list open issues carrying a `wip:*` label (number, title, holding ID); mark ones older than 6h\n  \"(expired, treated as free)\". Nothing else.\n- `unlock \u003Cissue#>` — show the issue's `wip:*`\u002F`in progress` labels, confirm, remove them.\n- a bare issue number — skip triage, show the issue and its comments, get the developer's read on it before writing\n  any code (plan mode \u002F a quick back-and-forth, not straight into the fix flow).\n- `stop` \u002F `drain` (as an argument, or said mid-run) — see \"Graceful stop\" below.\n- no argument — the full flow below (generate the instance ID first if concurrency applies).\n\n## Phase 0 — prerequisites and scope (counts only)\n\nList the repository labels first. Create the workflow label once with\n`gh label create \"in progress\" --color fbca04 --force`; claims depend on it and must not fail halfway through. Resolve\nthe two category labels from what is actually present: prefer `bug` for defects and `enhancement` for feature work,\nfall back to `Feature request` if that is the repository's convention, and ask when several plausible labels remain.\nReuse those exact names for the whole run.\n\nThen run `gh issue list --state open --limit 1000 --json number,title,labels,assignees,author`. Run the snooze wake pass\n(see below) and set aside issues with a foreign, non-expired `wip:*`. Report only counts — how many need categorizing,\nhow many bugs, how many feature requests, one line each for anything that woke up or is locked elsewhere. Do not\nenumerate the backlog; go straight into the per-issue walk. If the repository has more than 1,000 open issues, say\nthat this is the first batch rather than presenting the counts as complete.\n\n## Phase 1 — categorize\n\nUse the defect and feature labels resolved in Phase 0. Walk the uncategorized issues one at a time:\n\n1. A short summary (2–3 sentences) of what the issue wants.\n2. A weak title? Suggest a better one.\n3. Ask about **only this issue**, with a reasoned recommendation (bug vs. feature request; `question`\u002F`duplicate`\u002F\n   `wontfix` when in doubt), fold an optional title change into the same question. Wait for the answer.\n4. After the decision: claim, `gh issue edit N --add-label \"\u003Clabel>\"` (and `--title` if changed), release.\n\n## Phase 2 \u002F 3 — walk the bugs, then the feature requests\n\nOne at a time, waiting for the developer after each:\n\n1. `gh issue view N --comments` — explain the cause\u002Fcontext and give an honest read (worth fixing, effort, risk,\n   possible duplicate).\n2. If there are comments, summarize the thread — who said what, what's open. No comments → skip this step entirely.\n3. Ask what should happen, with a recommendation: **Fix** (below) \u002F **Ask the author** \u002F **Reject \u002F decline** \u002F\n   **Snooze** \u002F **Skip** (next issue, release any held lock).\n\nFor a feature request, \"Fix\" means \"implement,\" and scope is more often unsettled — clarify with the author or with\n`ha-planning` before code appears. Name dependencies between related requests when you see them.\n\n## Fix flow (issue approved)\n\n1. **Claim.**\n2. **Breaking-changes check first.** If the fix touches anything on `AGENTS.md`'s breaking-changes list (unique\n   IDs, entity IDs, entry data, state values, service signatures, removing a config option), stop and follow\n   [`ha-breaking-changes`](..\u002Fha-breaking-changes\u002FSKILL.md)'s warn-first procedure before writing any code — vutuv,\n   the source this skill generalizes from, has no equivalent concern, but this project does.\n3. **Isolate and branch.** Run `git status --short` before creating anything. If the current worktree is dirty, do\n   not absorb, stash, or discard those changes: create a separate worktree from `main`, or stop if an isolated\n   worktree is not available. In a clean tree, create a short kebab-case branch; never commit the fix on `main` in\n   this flow.\n4. **Implement.** If your agent runtime supports isolated sub-agent dispatch, use it; otherwise work directly on the\n   branch. For a bug, write the test that reproduces it first and watch it fail, then fix until it passes — a test\n   written after the fix may never have caught the bug. Other tests stay proportionate, per\n   [`ha-testing`](..\u002Fha-testing\u002FSKILL.md). If the fix turns out to need more than ~10 files or an architectural\n   change, stop and hand off to [`ha-planning`](..\u002Fha-planning\u002FSKILL.md) instead of continuing inline.\n5. **Gate once.** Run `script\u002Fcheck` (type-check + lint-check + spell-check) redirected to a log, judged by its exit\n   code — see [`blueprint-tooling`](..\u002Fblueprint-tooling\u002FSKILL.md)'s token-discipline note. Read the log, and only\n   the failing part, on a non-zero exit; iterate on just the failing piece, then re-run the full gate once more\n   before moving on.\n6. **Commit** using Conventional Commits\n   ([`blueprint.commit-message.instructions.md`](..\u002F..\u002Finstructions\u002Fblueprint.commit-message.instructions.md)).\n   **No manual version bump** — release-please derives it from the commit type once this lands on `main`\n   ([`ha-release`](..\u002Fha-release\u002FSKILL.md)).\n7. **Push and open the PR** (`gh pr create`, following `.github\u002Fpull_request_template.md` if the repo has one).\n   Open it ready for review, not as a draft: the gate above already passed, and CI skips drafts, so a draft would\n   leave the next step waiting for checks that never start.\n8. **Wait for CI:** `gh pr checks --watch --fail-fast`.\n   - Red → do not merge. Fix it and push again, watch again — or stop and report if it isn't yours to fix.\n   - Green → continue.\n9. **Merge**, per the autonomy rule at the top of this file. `gh pr merge \u003Cnr> --squash --delete-branch`.\n10. **Branch cleanup — every time, not optional.** `gh pr merge --delete-branch` normally removes both copies of the\n    branch. Refresh `main`, prune, and delete the local branch only if it still exists:\n\n    ```bash\n    git checkout main && git pull --ff-only\n    git fetch --prune origin\n    git show-ref --verify --quiet refs\u002Fheads\u002F\u003Cbranch> && git branch -D \u003Cbranch>\n    ```\n\n    The final command returning 1 because the branch is already gone is success, not a cleanup failure. If the fix\n    ran in an isolated worktree, remove that worktree before checking for the branch. Before moving on,\n    `git branch -vv` should show nothing marked `[origin\u002F\u003Cname>: gone]` — that marker is the leak.\n\n11. **Release the claim.** Remove `wip:\u003Cid>` and `in progress` even if the merge already closed the issue, then delete\n    the now-unused `wip:\u003Cid>` label definition.\n12. **Close the loop on the issue.** The merge may auto-close it (a `Fixes #N` PR body); either way, draft a short\n    shipped note — what now works, and why, in the language of the issue — per `AGENTS.md`'s \"Posting on the\n    developer's behalf\": get explicit sign-off on the draft unless this session is running in full-autonomy mode,\n    and disclose per `AI_POLICY.md` when it posts under the developer's identity without another marker.\n\n## Ask the author\n\nAlways an option in the per-issue walk. Get the author (`gh issue view N --json author`), draft a friendly question\nin their language, following the same posting-on-behalf policy as step 11 above. Claim → comment → release.\n\n## Reject \u002F decline\n\nWhen the developer decides an issue won't be done — out of scope, a design not wanted, a duplicate — close it\nkindly, never leave it to rot. Ask for the reason as part of the same question (a short recommendation is fine; \"no\nreason, just decline\" is a valid answer), then, after claiming: draft the closing note per the posting-on-behalf\npolicy, get sign-off, comment, `gh issue close N --reason \"not planned\"` (`--reason completed` only when it's\ngenuinely resolved another way), and release — a closed issue keeps no lock.\n\n## Snooze\n\nLabels `snoozed` + `snooze:YYYY-MM-DD` (name = wake date). After claiming: `gh label create snoozed --color c5def5\n--force`, `gh label create \"snooze:\u003Cdate>\" --color ededed --force`, apply both, release. At the start of every run,\ndo a wake pass: `snooze:\u003Cdate>` ≤ today → remove both labels, surface it at the top as woken up; delete now-empty\n`snooze:*` labels afterward. Bundle future ones into one line (\"N snoozed\").\n\n## Graceful stop (drain)\n\nThe developer can end a run at any point — not an emergency kill. **Finish work already in flight, start nothing\nnew:**\n\n1. Stop claiming, categorizing, commenting, snoozing, or dispatching anything new. The per-issue walk ends here.\n2. Let an in-flight fix finish: when it reports, apply the normal merge rule (green → merge + cleanup + release;\n   red\u002Funclear → hold the lock and report).\n3. For anything claimed but not yet started, list it and ask once whether to keep the claim for next time or release\n   it now (default recommendation: release the lock, keep any assignee).\n4. Report what shipped this run, what's still draining, and what was left untouched.\n",{"name":5,"description":7},"main","3a408474735fec0ac5d62b86fc842420a899e090",".agents\u002Fskills\u002Fha-issue-triage","https:\u002F\u002Fgithub.com\u002Famg0\u002Fha_ksenia","https:\u002F\u002Fgithub.com\u002Famg0","2026-11-15 01:01:02",{"source":8,"skillName":5,"commands":30,"prompt":40,"zipAvailable":41},[31,34,37],{"tool":32,"command":33},"npx","npx skills add amg0\u002Fha_ksenia --skill ha-issue-triage",{"tool":35,"command":36},"bunx","bunx skills add amg0\u002Fha_ksenia --skill ha-issue-triage",{"tool":38,"command":39},"pnpm","pnpm dlx skills add amg0\u002Fha_ksenia --skill ha-issue-triage","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Famg0\u002Fha_ksenia\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fha-issue-triage\u002FSKILL.md",true,[],[44,51,58,65,72,79],{"skillNo":45,"name":46,"nameZh":6,"description":47,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":48,"categories":49,"keywords":50,"publishedAt":18},"01M7AB5RYA92GRR64CP6VMWV6Z","issues","Triage\u002Ffix the GitHub issue backlog via branch + PR, with auto-merge on green CI for this session.","https:\u002F\u002Fgithub.com\u002Famg0\u002Fha_ksenia\u002Fblob\u002Fmain\u002F.github\u002Fskills\u002Fissues\u002FSKILL.md",[],[],{"skillNo":52,"name":53,"nameZh":6,"description":54,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":55,"categories":56,"keywords":57,"publishedAt":18},"01M7AB5G6DR1KGAPRT7JHDCD2Y","ha-translations","Add or fix user-facing strings for this Home Assistant custom integration — translations\u002Fen.json and other languages, icons.json, entity names and states, config flow labels and errors, service action descriptions, exception messages, and repair issue text. Use when asked to \"update translations\", \"add a translation key\", \"translate\", \"fix the entity name\", \"add strings for the new sensor\u002Faction\u002Foption\", \"add icons\", or when hassfest reports a missing or unused translation key. Covers the key structure, which keys each feature needs, placeholders, the icons.json contract, and the project's strict rule about only touching en.json. SYMPTOMS — load this if you are about to: write user-facing English into Python; edit a language file other than `en.json`; add an entity without its `entity.\u003Cplatform>.\u003Ckey>.name` key; skip `data_description` on a config flow field; or set `icon=` instead of using `icons.json`.","https:\u002F\u002Fgithub.com\u002Famg0\u002Fha_ksenia\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fha-translations\u002FSKILL.md",[],[],{"skillNo":59,"name":60,"nameZh":6,"description":61,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":62,"categories":63,"keywords":64,"publishedAt":18},"01M7AB56JBRMF19W4N2NG9Y5GW","ha-testing","Write, fix, or extend automated tests for this Home Assistant custom integration using pytest and pytest-homeassistant-custom-component. Use when asked to \"add tests\", \"write a test\", \"the tests fail\", \"improve coverage\", \"add a regression test\", \"test the config flow\", \"update snapshots\", or when a behavioural change needs verification before it can be called done. Covers the tests\u002F layout, the conftest.py fixtures this project needs, mocking the API client, config flow and entity test patterns, registry assertions, syrupy snapshots, freezing time, and the script\u002Ftest commands. SYMPTOMS — load this if you are about to: instantiate an entity class directly instead of asserting on `hass.states`; write a test without `enable_custom_integrations`; use `time.sleep` instead of `async_fire_time_changed`; silence a DeprecationWarning to make the suite pass; or commit a snapshot you have not read.","https:\u002F\u002Fgithub.com\u002Famg0\u002Fha_ksenia\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fha-testing\u002FSKILL.md",[],[],{"skillNo":66,"name":67,"nameZh":6,"description":68,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":69,"categories":70,"keywords":71,"publishedAt":18},"01M7AB4V9XKDYRDCRHEA68GRZ5","ha-service-action","Add, change, or remove a Home Assistant service action (also called a \"service\" or \"action\") in this custom integration. Use when asked to \"add a service\", \"add an action\", \"expose X as a service call\", \"add a service that returns data\", \"make this callable from an automation\", or when editing services.yaml, the service_actions\u002F package, service schemas, service response data, or the async_setup() registration. Covers the services.yaml contract, probatio schemas, selectors, entity\u002Fdevice targeting, response data, translated error messages, and the Bronze quality-scale rule that actions must be registered in async_setup(). SYMPTOMS — load this if you are about to: register an action in `async_setup_entry()`; add a `services.yaml` field without a selector; add an `entity_id` field where `target:` belongs; raise a plain English `HomeAssistantError`; or swallow an action failure into a log line.","https:\u002F\u002Fgithub.com\u002Famg0\u002Fha_ksenia\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fha-service-action\u002FSKILL.md",[],[],{"skillNo":73,"name":74,"nameZh":6,"description":75,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":76,"categories":77,"keywords":78,"publishedAt":18},"01M7AB4J65REZ9T9VS9MQH73CC","ha-release","Cut a release of this Home Assistant custom integration, or prepare the commits that feed it. Use when asked to \"release\", \"cut a version\", \"bump the version\", \"prepare release notes\", \"what will the next version be\", \"write a commit message\", \"why didn't release-please pick this up\", or when reviewing whether a change is user-facing. Covers the Conventional Commits rules commitlint enforces here, how release-please derives the version and changelog, the pre-1.0 bump rules, script\u002Fversion and script\u002Frelease-notes, skipping a commit from the notes, and the commit rules this project enforces for agents. SYMPTOMS — load this if you are about to: commit without an explicit request; offer to `git push`; label a user-facing fix as `chore` (chore, like every hidden type, never reaches the changelog and never triggers a release); write a commit subject over 100 characters or in sentence case; or edit the version in `manifest.json` by hand.","https:\u002F\u002Fgithub.com\u002Famg0\u002Fha_ksenia\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fha-release\u002FSKILL.md",[],[],{"skillNo":80,"name":81,"nameZh":6,"description":82,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":83,"categories":84,"keywords":85,"publishedAt":18},"01M7AB48WMPGDK2REBC3WS8T32","ha-quality-review","Perform a structured quality review of this Home Assistant custom integration, either as a full audit or scoped to a diff. Use when asked to \"review the integration\", \"audit the code\", \"check quality scale\", \"are we bronze\u002Fsilver\u002Fgold\", \"is this ready to release\", \"review my changes\", \"what's missing before publishing\", or before opening a pull request that touches integration code. Covers running the automated gates first, then auditing architecture, the Integration Quality Scale rules, error handling, security, performance, user experience, and documentation — and reporting findings ranked by severity with concrete fixes. SYMPTOMS — load this if you are about to: report a review that only restates linter output; claim real-device testing that did not happen; give a finding without a file path and a fix; or judge quality without running `script\u002Fhassfest` and the test suite first.","https:\u002F\u002Fgithub.com\u002Famg0\u002Fha_ksenia\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fha-quality-review\u002FSKILL.md",[],[]]