[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M74FTCM43KBGRC9M39SS7Z1A":3,"skill-same:01M74FTCM43KBGRC9M39SS7Z1A":49},{"skillNo":4,"name":5,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":7,"ownerLogin":8,"ownerAvatarUrl":9,"ownerAccountNo":10,"ownerDisplayName":11,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":15,"categories":16,"keywords":17,"publishedAt":18,"linkOnly":19,"markdownContent":20,"markdownBody":20,"frontMatter":21,"refName":22,"commitSha":23,"skillDirectory":24,"forksCount":13,"repositoryHtmlUrl":25,"repositoryHomepageUrl":6,"ownerHtmlUrl":26,"ownerBlogUrl":6,"ownerDescription":27,"updatedAt":28,"install":29,"resources":42},"01M74FTCM43KBGRC9M39SS7Z1A","finish-task",null,"Astrumon\u002Fspovishun-skills","Astrumon","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F67917191?v=4","01M74FSGPFMM3XWY2PMSEN8Z2W","Danylo Bidnyk","01M74FSGQ12JX4QM4FEY6NEH6N",0,"MIT","https:\u002F\u002Fgithub.com\u002FAstrumon\u002Fspovishun-skills\u002Fblob\u002Fmain\u002Fskills\u002Ffinish-task\u002FSKILL.md",[],[],"2026-11-11 15:11:14",false,"# Finish Task\n\nRuns the task-completion gate symmetric to the \"start new task\" flow: a **non-blocking**\ncomment-hygiene pass, a **blocking** quality gate (tests → build → lint), then an\n**advisory** code review on the task diff,\nbefore offering push \u002F PR \u002F Notion status update. Closes with a plain-language summary in\nthe project's own language. Never auto-merges and never sets a task to `Done` automatically.\n\n## Workflow\n\nExecute the steps **in order**. Do not skip ahead — a failure in the blocking gate stops\nthe whole flow.\n\n### Step 1: Resolve the active task\n\n**1a.** Get the current branch:\n```\ngit rev-parse --abbrev-ref HEAD\n```\nIf the branch is `{{GIT_DEVELOP_BRANCH}}` (default `develop`) or `main`, stop: there is no\nactive task to finish — ask the user to switch to a task branch first.\n\n**1b.** Derive the cache folder: replace `\u002F` with `-` in the branch name and append `_prd`\n(e.g. `feature\u002F{{PROJECT_PREFIX}}-77-foo` → `feature-{{PROJECT_PREFIX}}-77-foo_prd`).\n\n**1c.** `Read` `.dev-context\u002F{folder}\u002Ftask.json` (if present) for the task **Goal** and\n**Definition of Done** — use them later to confirm the work matches what was asked. If the\nfile is absent, say so once and continue, but do **not** claim in Step 7 that the Definition\nof Done was verified.\n\n**1d.** Check the tree is clean:\n```\ngit status --porcelain\n```\nNon-empty output → **warn before continuing**: Step 3 runs the gate on the **working tree**,\nwhile Step 5 reviews only `git diff {{GIT_DEVELOP_BRANCH}}...HEAD` — the **committed** diff.\nAnything left uncommitted therefore passes the gate unreviewed and stays out of the PR. List\nthe dirty paths and offer to land them first: `\u002Fcommit-sections` for a large change that wants\nslicing, `\u002Fcommit` for one coherent change. Continue only if the user says so, and in that case\nsay in Step 7 that the review covered the committed diff only.\n\n### Step 2: Resolve the project commands\n\nRead the build\u002Ftest\u002Flint commands from the consumer's `CLAUDE.md` **`## Commands`** and\n**`## Testing`** sections — do **not** hardcode a build tool. Identify:\n- the **test** command(s),\n- the **build** command,\n- the **blocking** linter\u002Fformatter check(s),\n- any **non-blocking** static analysis the project documents (e.g. a CI gate marked\n  non-blocking).\n\n> Example (Spovishun): tests `.\u002Fgradlew test` (+ `.\u002Fgradlew integrationTest`), build\n> `.\u002Fgradlew build`, blocking lint `.\u002Fgradlew ktlintCheck`, non-blocking `.\u002Fgradlew detekt`.\n\nIf `CLAUDE.md` documents no commands, **stop and ask the user** which test, build and lint\ncommands to run. Do not infer a build tool from the file tree — a guessed command that\nsilently does nothing reports a green gate on unverified code.\n\n### Step 2b: Comment hygiene (non-blocking)\n\nIf the `comment-hygiene` skill is available (same presence check as Step 5b), invoke it scoped\nto the task diff `{{GIT_DEVELOP_BRANCH}}...HEAD`. If it is not available, skip this step\nsilently.\n\nIt proposes comment edits as a diff and applies them only after the user confirms. Nothing\nhere stops the flow: a declined proposal, an empty one or a file it restored all continue to\nStep 3.\n\nIt runs **before** the gate on purpose: Step 3 runs on the working tree, so a removal that\nleaves a blank line ktlint rejects is caught there. Applied edits stay **uncommitted** — they\nare outside the committed diff Step 5 reviews and outside the PR — so right after they are\napplied, offer `\u002Fcommit` to land them. Never commit them yourself.\n\n### Step 3: Blocking gate (tests → build → lint), in order\n\nRun, in this exact order:\n1. **Tests** — including integration tests if the project defines them.\n2. **Build**.\n3. **Blocking lint\u002Fformat check**.\n\nOn the **first failure**: STOP the gate. Surface the failing output as `file:line` references\nand a short diagnosis. Do **not** run the code review, do **not** offer push\u002FPR, do **not**\ntouch Notion — skip Steps 4, 5 and 6 entirely. Then **jump straight to Step 7** and emit the\nsummary, marking the review as not run. After the summary, hand control back so the user can\nfix and re-run `finish task`.\n\n### Step 4: Non-blocking static analysis\n\nRun any non-blocking analysis (e.g. detekt). **Report** findings, but never fail the gate on\nthem.\n\n### Step 5: Advisory code review on the task diff\n\nCompute the task diff against the base branch:\n```\ngit diff {{GIT_DEVELOP_BRANCH}}...HEAD\ngit diff --numstat {{GIT_DEVELOP_BRANCH}}...HEAD\n```\n(If `{{GIT_DEVELOP_BRANCH}}` is empty, default to `develop`.)\n\nThe review is not one-size-fits-all: classify what the diff touched first (5a), resolve at\nmost three domain specialists that are actually available (5b), then run one review (5c).\n\n#### Step 5a: Classify the diff\n\n`--numstat` gives per-file volume; the hunks give content signals. Match every changed file\nagainst this table. Path\u002Fextension signals come first, content signals confirm.\n\n| Domain | Path \u002F extension signals | Content signals |\n|---|---|---|\n| `ui-compose` | `*.kt` under `ui\u002F`, `presentation\u002F`, `composeApp\u002F` | `@Composable`, `androidx.compose.`, `Modifier`, `remember`, `LaunchedEffect` |\n| `concurrency` | any `*.kt` | `suspend fun`, `launch {`, `async {`, `Flow\u003C`, `withContext`, `CoroutineScope`, `Dispatchers.`, `CancellationException` |\n| `persistence` | `data\u002F`, `db\u002F`, `repository\u002F`; `*.sq`, `schemas\u002F*.json` | `: Table`, `transaction {`, `safeDbQuery`, Exposed \u002F SQLDelight \u002F Room imports, `@Entity` |\n| `sql-schema` | `db\u002Fmigration\u002F*.sql`, any `*.sql` | `CREATE INDEX`, `ALTER TABLE`, `JOIN`, `EXPLAIN` |\n| `di` | `di\u002F`, `*Module.kt` | `module {`, `single\u003C`, `factory\u003C`, `viewModelOf`, `by inject` |\n| `bot` | `presentation\u002Fbot\u002F`, `*Handler.kt` | Telegram SDK imports, `command(`, `InlineKeyboard` |\n| `networking` | `network\u002F`, `api\u002F`, `*Client.kt` | `HttpClient`, `expectSuccess`, Ktor client imports |\n| `tests` | `src\u002Ftest\u002F`, `src\u002F*Test\u002F`, `commonTest`, `*Test.kt`, `*Spec.kt` | `@Test`, `runTest`, `mockk`, `Turbine` |\n| `build` | `build.gradle.kts`, `settings.gradle.kts`, `gradle\u002Flibs.versions.toml`, `gradle.properties` | — (path is enough) |\n| `containers` | `Dockerfile*`, `docker-compose*.y*ml`, `.dockerignore` | — (path is enough) |\n\n**Trigger threshold.** Count only **substantive** changed lines. Excluded: `import` \u002F\n`package` lines, blank lines, comment-only lines, and lines whose only change is formatting\nor a rename.\n\nA domain is **touched** when it has **≥ 10 substantive changed lines**, or **≥ 2 files** with\nany substantive change.\n\n*Worked counter-example:* one changed `import androidx.compose.foundation.layout.Column`\nline does **not** make `ui-compose` touched.\n\n**Exception — always touched at any size:** a new or modified DB migration, a `Dockerfile` \u002F\ncompose file, or a version-catalog entry. A schema, an image and a dependency version are\nsubstantive at one line.\n\nA file may belong to more than one domain (a `Repository.kt` full of `suspend fun` counts for\nboth `persistence` and `concurrency`). Count it in each.\n\n#### Step 5b: Rank, cap, and resolve by presence\n\n1. Rank touched domains by **share of substantive changed lines**, descending.\n2. **Cap: at most 3** specialists per run, at most **one per domain row**. Beyond that the\n   review context balloons and `code-reviewer` drowns in it.\n3. For each of the top domains, walk its candidate list in order and take the **first one\n   available in this session**:\n\n| Domain | Candidates (first available wins) |\n|---|---|\n| `ui-compose` | `compose-multiplatform` → `kmp-multiplatform-specialist` |\n| `concurrency` | `kotlin-coroutines-expert` * → `kotlin-specialist` |\n| `persistence` | `postgresql-exposed-orm` → `kmp-persistence` → `database-optimizer` |\n| `sql-schema` | `database-optimizer` → `database-reviewer` (agent) |\n| `di` | `dependency-injection-architecture` → `koin-kmp` |\n| `bot` | `telegram-bot-development` |\n| `networking` | `ktor-client-kmp` |\n| `tests` | `unit-testing-kotlin` → `kmp-testing` |\n| `build` | `gradle-build-auditor` |\n| `containers` | `docker-deployment` |\n| *Kotlin general* — only if **no** row above resolved and the diff is Kotlin-heavy | `kotlin-code-review` * → `thermo-nuclear-code-quality-review` * → `kotlin-reviewer` (agent) |\n\n`*` = not shipped with this skill. Reachable only when the consumer happens to have it in\n`~\u002F.claude\u002Fskills\u002F` or via another plugin.\n\n**Availability check — resolve by presence, never by assumption.** This skill installs into\nprojects that have none of the skills above on disk. Never assume a row's target exists.\n\n1. Primary: the id appears in the list of skills \u002F agents available in this session. That\n   list is the only check that sees project skills, user-level `~\u002F.claude\u002Fskills\u002F`, and\n   plugin skills alike.\n2. Fallback when that is unclear: `Glob` for `.claude\u002Fskills\u002F\u003Cid>\u002FSKILL.md`,\n   `~\u002F.claude\u002Fskills\u002F\u003Cid>\u002FSKILL.md`, `.claude\u002Fagents\u002F\u003Cid>\u002FAGENT.md`.\n\n**If no candidate in a row is available, skip the row silently** — no error, no warning, no\nmention in the report. A project with none of them resolves zero rows, and Step 5 degrades\nto a plain `code-reviewer` pass. That is a correct outcome, not a degraded one.\n\n#### Step 5c: Run the review\n\n1. Invoke each resolved specialist **scoped to its own domain's files only**, and ask it for\n   **domain findings as input** — not for a report of its own.\n2. Then invoke the `code-reviewer` skill **once**, scoped to the full diff, passing the\n   specialist findings as extra context.\n3. Emit `code-reviewer`'s Critical \u002F Major \u002F Minor + Verdict report. Above it, one line names\n   which specialists ran, e.g. `Domain review: compose-multiplatform, kotlin-specialist`.\n   Skipped rows are simply absent from that line; if none resolved, omit the line.\n\n- **Critical** findings must be surfaced explicitly and **acknowledged by the user** before\n  offering `Done` — but they do **not** hard-block the flow.\n- Major \u002F Minor are advisory.\n\n### Step 6: On green — offer next actions\n\nSummarize the gate result and the review verdict, then offer (do not perform automatically):\n- **push** the branch,\n- **open a PR**,\n- set Notion `Status = Done` — **only after the PR is merged** (per the project's Task Status\n  Workflow).\n\nNever auto-merge. Never set `Done` automatically. Confirm the work satisfies the task's\nDefinition of Done from Step 1 before recommending completion.\n\n### Step 7: Plain-language summary in {{PROJECT_LANGUAGE}}\n\nAlways runs — on green after Step 6, and immediately after a Step 3 blocking failure. It sits\n**alongside** the technical report from Step 5; it never replaces it.\n\n**Brief:**\n\n> Write the summary in the project language `{{PROJECT_LANGUAGE}}` (ISO code — `uk` is\n> Ukrainian, `en` is English), in plain technical prose: short sentences, one idea per\n> sentence, no metaphors.\n>\n> **Technical terms stay in their original form.** Do not translate `coroutine`, `Flow`,\n> `CancellationException`, `detekt`, `ktlint`, `Composable`, `migration`, `PR`. What gets\n> simplified is **sentence structure, not vocabulary** — this is not ELI5.\n>\n> If the project defines a ubiquitous language (a `CONTEXT.md`, or a Glossary section in the\n> root `CLAUDE.md`), use its terms.\n>\n> **Under 250 words.**\n\nFour sections, with the headings themselves written in `{{PROJECT_LANGUAGE}}` (canonical\nEnglish names given here for reference):\n\n1. **What changed** — 2–4 sentences derived from the diff.\n2. **What the blocking gate reported** — tests \u002F build \u002F lint: passed, or the first failure\n   and where it is.\n3. **What the review found** — counts by severity plus the worst finding. State `not run`\n   when Step 5 was skipped, and say that it covered the committed diff only when Step 1d\n   found a dirty tree and the user continued anyway.\n4. **What is left for you to do** — concrete next actions: fix X, push, open a PR,\n   acknowledge a Critical finding, set Notion `Done` after the merge. If Step 2b applied\n   comment edits that are still uncommitted, say so here.\n\n## Review boundaries\n\nWho writes what, so nothing is reviewed twice:\n\n- **`code-reviewer`** — always runs, and is the **sole author** of the Critical \u002F Major \u002F\n  Minor + Verdict report.\n- **Specialist skills from Step 5b** — add domain depth and feed their findings into\n  `code-reviewer`. They never emit their own severity block or verdict.\n- **`two-axis-code-review`** — a *different* review (Standards vs Spec) that the user invokes\n  directly. `finish-task` **never** invokes it: running both would put two verdicts on one\n  diff.\n- **`kotlin-reviewer` \u002F `code-architecture-reviewer` \u002F `database-reviewer` agents** — deep\n  standalone audits. At most one is reached, and only through the last-resort row in the\n  table. Its output is folded into `code-reviewer`'s report, never printed separately.\n\n## Do NOT\n\n- Do **not** make the routing table safe by narrowing this skill to a single stack. The\n  presence check in Step 5b is what makes the table safe; a project with none of those skills\n  must still get the full gate.\n- Do **not** invoke a specialist skill without checking it is available first.\n- Do **not** let a specialist skill write its own Critical\u002FMajor\u002FMinor report.\n- Do **not** skip Step 7 when the blocking gate fails — that is exactly when the summary is\n  most useful.\n- Do **not** push, open a PR, merge, or set Notion `Status = Done` without the user asking.\n- Do **not** commit the Step 2b comment edits, and do **not** let that pass change code — it\n  is comments-only, and it is not a review axis.\n\n## Example run\n\nUser is on `feature\u002F{{PROJECT_PREFIX}}-142-schedule-digest`. The diff against\n`{{GIT_DEVELOP_BRANCH}}` is 180 substantive lines: ~120 in `domain\u002Fschedule\u002F` (all\n`suspend fun` and `Flow\u003C`), ~45 in `data\u002FScheduleRepository.kt` (Exposed `Table` + a new\nmigration), 15 in `ScheduleModule.kt`.\n\nExpected outcome:\n\n- Step 3 runs tests → build → lint; all green.\n- Step 5a marks `concurrency`, `persistence` and `di` as touched. `sql-schema` is touched too\n  (a migration is substantive at any size).\n- Step 5b ranks them, caps at 3, and resolves `kotlin-specialist`, `postgresql-exposed-orm`\n  and `database-optimizer`. `di` falls outside the cap. In a project where none of those\n  three is installed, all rows are skipped and nothing is said about it.\n- Step 5c prints `Domain review: kotlin-specialist, postgresql-exposed-orm, database-optimizer`\n  and then one `code-reviewer` report with a single Verdict.\n- Step 6 offers push \u002F PR \u002F Notion `Done`, performing none of them.\n- Step 7 prints the four-section summary in `{{PROJECT_LANGUAGE}}`, under 250 words, with\n  `Flow`, `suspend`, `migration` and `ktlint` left untranslated.\n\n## Related Skills\n\n- `comment-hygiene` — the comment pass invoked in Step 2b.\n- `code-reviewer` — the review pass invoked in Step 5c; owns the report format.\n- `commit` — committing the work before the gate runs.\n- `commit-sections` — slicing a large finished change into ordered commits before the gate\n  runs; the counterpart Step 1d points at.\n- `git-workflow-pr-writing` — PR body and branch conventions for Step 6.\n",{},"main","4ad702b3d4f403306c1a016277020693dd669f55","skills\u002Ffinish-task","https:\u002F\u002Fgithub.com\u002FAstrumon\u002Fspovishun-skills","https:\u002F\u002Fgithub.com\u002FAstrumon","Software Engineer","2026-10-14 01:01:46",{"source":7,"skillName":5,"commands":30,"prompt":40,"zipAvailable":41},[31,34,37],{"tool":32,"command":33},"npx","npx skills add Astrumon\u002Fspovishun-skills --skill finish-task",{"tool":35,"command":36},"bunx","bunx skills add Astrumon\u002Fspovishun-skills --skill finish-task",{"tool":38,"command":39},"pnpm","pnpm dlx skills add Astrumon\u002Fspovishun-skills --skill finish-task","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FAstrumon\u002Fspovishun-skills\u002Fblob\u002Fmain\u002Fskills\u002Ffinish-task\u002FSKILL.md",true,[43],{"path":44,"resourceType":45,"resourceTypeDesc":46,"sizeBytes":47,"sourceHtmlUrl":48},"manifest.yaml","REFERENCE","参考文档",934,"https:\u002F\u002Fgithub.com\u002FAstrumon\u002Fspovishun-skills\u002Fblob\u002Fmain\u002Fskills\u002Ffinish-task\u002Fmanifest.yaml",[50,56,62,69,74,79],{"skillNo":51,"name":52,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":7,"ownerLogin":8,"ownerAvatarUrl":9,"ownerAccountNo":10,"ownerDisplayName":11,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":53,"categories":54,"keywords":55,"publishedAt":18},"01M74FX71JH2KNG5KX7TZHSKCB","universal-skill","https:\u002F\u002Fgithub.com\u002FAstrumon\u002Fspovishun-skills\u002Fblob\u002Fmain\u002Ftest\u002Ffixtures\u002Fsource\u002Fskills\u002Funiversal-skill\u002FSKILL.md",[],[],{"skillNo":57,"name":58,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":7,"ownerLogin":8,"ownerAvatarUrl":9,"ownerAccountNo":10,"ownerDisplayName":11,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":59,"categories":60,"keywords":61,"publishedAt":18},"01M74FX57GTEFPJVCPWPSP6CCW","notion-skill","https:\u002F\u002Fgithub.com\u002FAstrumon\u002Fspovishun-skills\u002Fblob\u002Fmain\u002Ftest\u002Ffixtures\u002Fsource\u002Fskills\u002Fnotion-skill\u002FSKILL.md",[],[],{"skillNo":63,"name":64,"nameZh":6,"description":65,"descriptionZh":6,"repositoryFullName":7,"ownerLogin":8,"ownerAvatarUrl":9,"ownerAccountNo":10,"ownerDisplayName":11,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":66,"categories":67,"keywords":68,"publishedAt":18},"01M74FX30F3K8906P5ASQ3DNXR","inline-frontmatter-skill","Hand-authored description that already lives in the body and must not be double-wrapped.","https:\u002F\u002Fgithub.com\u002FAstrumon\u002Fspovishun-skills\u002Fblob\u002Fmain\u002Ftest\u002Ffixtures\u002Fsource\u002Fskills\u002Finline-frontmatter-skill\u002FSKILL.md",[],[],{"skillNo":70,"name":52,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":7,"ownerLogin":8,"ownerAvatarUrl":9,"ownerAccountNo":10,"ownerDisplayName":11,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":71,"categories":72,"keywords":73,"publishedAt":18},"01M74FX18AGHM2KHQ502PMJZRJ","https:\u002F\u002Fgithub.com\u002FAstrumon\u002Fspovishun-skills\u002Fblob\u002Fmain\u002Ftest\u002Ffixtures\u002Fsource-v2\u002Fskills\u002Funiversal-skill\u002FSKILL.md",[],[],{"skillNo":75,"name":58,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":7,"ownerLogin":8,"ownerAvatarUrl":9,"ownerAccountNo":10,"ownerDisplayName":11,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":76,"categories":77,"keywords":78,"publishedAt":18},"01M74FWZJ193G810P0ZJ1PKEZ9","https:\u002F\u002Fgithub.com\u002FAstrumon\u002Fspovishun-skills\u002Fblob\u002Fmain\u002Ftest\u002Ffixtures\u002Fsource-v2\u002Fskills\u002Fnotion-skill\u002FSKILL.md",[],[],{"skillNo":80,"name":81,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":7,"ownerLogin":8,"ownerAvatarUrl":9,"ownerAccountNo":10,"ownerDisplayName":11,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":82,"categories":83,"keywords":84,"publishedAt":18},"01M74FWPNEZWWGQ2Q37N6MMWP1","write-a-skill","https:\u002F\u002Fgithub.com\u002FAstrumon\u002Fspovishun-skills\u002Fblob\u002Fmain\u002Fskills\u002Fwrite-a-skill\u002FSKILL.md",[],[]]