[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M77HCXSTV37RDQMKJ6EB5SPP":3,"skill-same:01M77HCXSTV37RDQMKJ6EB5SPP":47},{"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":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":24,"commitSha":25,"skillDirectory":26,"forksCount":27,"repositoryHtmlUrl":28,"repositoryHomepageUrl":29,"ownerHtmlUrl":30,"ownerBlogUrl":31,"ownerDescription":6,"updatedAt":32,"install":33,"resources":46},"01M77HCXSTV37RDQMKJ6EB5SPP","dos-dispatch-loop",null,"Run recurring `dos-dispatch` cycles, switching to `dos-replan` when the backlog drains and stopping on the kernel's loop verdict. Use for unattended dispatch->replan->dispatch work across disjoint lanes.","anthony-chaudhary\u002Ffak","anthony-chaudhary","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F18080164?v=4","01M77GSKCAEKV3JS07J40B94DY","Anthony Chaudhary","01M77GSKD26Z7F4FQQD4VA30VX",48,"Apache-2.0","https:\u002F\u002Fgithub.com\u002Fanthony-chaudhary\u002Ffak\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fdos-dispatch-loop\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: dos-dispatch-loop\ndescription: \"Run recurring `dos-dispatch` cycles, switching to `dos-replan` when the backlog drains and stopping on the kernel's loop verdict. Use for unattended dispatch->replan->dispatch work across disjoint lanes.\"\n---\n\n# dos-dispatch-loop — the generic dispatch⇄replan cadence\n\n> **The unattended plan-and-ship loop.** It runs `\u002Fdos-dispatch` repeatedly and\n> falls to `\u002Fdos-replan` when a lane drains, stopping on a typed, kernel-decided\n> condition (not a prose guess). The stop\u002Fcontinue logic is the kernel's\n> `loop_decide.decide` — the loop carries counters, the kernel decides. Several\n> loops on disjoint lanes run in parallel, each holding its own lane lease.\n\nThe stop conditions are the kernel's, in one place:\n\n1. **iteration cap** — reached `max_iterations` (default 10).\n2. **drained-twice** — a DRAIN after a *productive* `\u002Fdos-replan` that itself\n   followed a DRAIN (the lane is genuinely exhausted).\n3. **consecutive-unclear** — the dispatch subprocess is failing systematically.\n4. **rate-limited** — a usage window is exhausted (don't burn launches).\n5. **launch-failed** — a subprocess never started.\n6. **pick-held-invariant** — the next unit is held ONLY by a reason a re-dispatch\n   cannot change (draft-class \u002F operator-gated \u002F soak-open \u002F dependency-unmet);\n   re-dispatching it would re-block identically, so honest-STOP + surface the hold.\n7. **pick-cooldown** — the next unit was attempted-and-didn't-move inside its\n   cooldown window AND nothing fresher is offerable; re-dispatching it would\n   re-storm a known drain (the ~5%-shipping re-pick loop the bare loop hit).\n8. **not-ratcheting** — the OUTER RATCHET (docs\u002F351): the loop has run too many\n   iterations in a row with no *witnessed net gain* — the reconcile-VERIFIED\n   ship-count (Step 3) is not rising, even though each iteration reported SHIPPED.\n   The loop is running but not improving (the \"spinning, narrating progress while\n   net-shipping nothing real\" failure that worsens the longer the loop runs). The\n   verdict is the kernel's `improve` ESCALATE, read VERBATIM — not the loop's\n   self-report. Hand the judgment back to a human.\n\nConditions 6–7 are the docs\u002F207 anti-churn rungs (the loop stops re-picking work\nit cannot move); condition 8 is the docs\u002F351 ratchet (the loop stops when it is\nmoving but not *improving* — RSI made first-class here, the same `improve` keep-gate\nthe `\u002Fdos-self-improve` loop uses, now gating each dispatch iteration's net gain).\n\n## Inputs\n\n- `--lane \u003Cname>` (optional) — focus the whole loop on one lane (fixed for the\n  run; a bare loop auto-picks a free lane at Step 0).\n- `--gate hard|soft|drive` (default `hard`) — the verdict policy. `hard` routes a\n  non-LIVE verdict through `\u002Fdos-replan`; `soft`\u002F`drive` stop on a true DRAIN;\n  `drive` self-heals a STALE-STAMP inline.\n- `--max-iterations \u003CN>` (default 10).\n\n## Step 0 — Pre-flight: take the lane, read the taxonomy\n\n```bash\ndos doctor --workspace . --json\npython tools\u002Fdos_fleet_lease.py --workspace . acquire --lane \u003CLANE> --kind cluster\n```\n\nThe fleet lease wrapper materializes the live lease union, feeds it to the kernel\nacquire, then republishes this node's held set. For ordinary shard lanes this is the\nsame-host fast path; for the three global lanes (`abi`, `release`, `global`) it uses\nthe GitRefStore cross-node transport by default (#21). A REFUSE means a sibling loop\nalready holds an overlapping lane — pick a free one from `kernel.free_clusters` or\nexit. Initialise the loop counters (iteration=1, the breakers at 0).\n\n> **`dos lease-lane acquire` is the lease AUTHORITY; `dos arbitrate` is ADVISORY.**\n> Only `acquire` folds the durable lane-journal WAL under the lease mutex and\n> journals the grant, so only its verdict binds. Honor its REFUSE; never start\n> editing a tree on the strength of an arbitrate GO alone (arbitrate holds no lock,\n> so even a correct GO can be stale by acquire time).\n>\n> The advisory has two surfaces and they do **not** behave the same:\n>\n> - `dos arbitrate` (the **CLI**) reads the live WAL by default, so it agrees with\n>   `acquire`. Passing `--leases '[]'` opts out into an empty world.\n> - The **MCP tool** `dos_arbitrate` reads **nothing** — it arbitrates only against\n>   the `live_leases` you hand it, and an omitted argument means \"the world is\n>   empty\", so every lane reads free. Observed live: MCP `dos_arbitrate` answered\n>   `\"cluster lane 'tools' free — admitted.\"` for a lane `dos lease-lane acquire`\n>   refused the same second as `\"already held by a live loop\"`\n>   ([dos-kernel#246](https:\u002F\u002Fgithub.com\u002Fanthony-chaudhary\u002Fdos-kernel\u002Fissues\u002F246)).\n>   Until that lands, either use the CLI form or pass the real set explicitly\n>   (`dos lease-lane live` \u002F `fak leaseref live`). A FREE from a checker that cannot\n>   see the store is not evidence of a free lane.\n\n## Step 1 — Pick-selection: skip held + cooled units (the anti-churn gate)\n\nBefore a `dispatch` iteration offers a unit, screen the candidate unit set so the\nloop never re-storms work it cannot move (the docs\u002F207 §6 throughline). For each\ncandidate, in order:\n\n```bash\ndos pickable \u003CUNIT> --state '\u003Chost-gathered state>'   # OFFERABLE=0; HELD=per-reason code\ndos cooldown \u003CUNIT>                                    # CLEAR=0; RECENTLY_ATTEMPTED=3\n```\n\n- A `pickable` exit of **0** (OFFERABLE) **and** a `cooldown` exit of **0** (CLEAR)\n  → this unit is dispatchable; offer it and proceed to the iteration.\n- A `pickable` HELD by a re-dispatch-CURABLE reason (IN_FLIGHT \u002F SOFT_CLAIMED \u002F\n  STALE_CLAIM \u002F UNPARSEABLE) or a `cooldown` of **3** (RECENTLY_ATTEMPTED) → **skip\n  this unit, try the next candidate** (the skip-to-next is pick-selection's job).\n- A `pickable` HELD by a re-dispatch-INVARIANT reason (DRAFT_CLASS=10 \u002F\n  OPERATOR_GATED=11 \u002F SOAK_OPEN=12 \u002F DEPENDENCY_UNMET=13) → carry that verdict into\n  Step 2; the kernel will honest-STOP on it (don't re-dispatch a unit a re-dispatch\n  cannot un-gate).\n\nWhen EVERY remaining candidate is skipped (all cooled \u002F curably-held), carry the\nlast `cooldown` RECENTLY_ATTEMPTED (or the invariant `pickable` hold) into Step 2\nas the loop's pre-dispatch evidence — the kernel turns \"nothing fresh is offerable\"\ninto the `pick-cooldown` \u002F `pick-held-invariant` honest-STOP. **Do not re-dispatch\nthe cooled\u002Fheld unit yourself** — that is the re-pick storm this gate prevents.\n\n## Step 1b — Run one iteration\n\nFor a `dispatch` iteration, invoke `\u002Fdos-dispatch --lane \u003CLANE>` (it snapshots,\ngates, and ships). For a `replan` iteration, invoke `\u002Fdos-replan`. Capture the\niteration's outcome:\n\n- a **dispatch** iteration that reached the gate carries a typed verdict — get it\n  from `dos gate` over the packet's dispositions sidecar (LIVE\u002FDRAIN\u002FSTALE-STAMP\u002F\n  BLOCKED\u002FRACE).\n- a **replan** iteration carries a productivity signal (did it refill\u002Fgarden?).\n\n## Step 2 — Decide continue \u002F replan \u002F stop (the kernel decides)\n\nThis is the load-bearing step: **the decision is a kernel mechanism, not prose.**\nFeed the iteration outcome + the carried counters to the loop decider. In code a\nhost calls `dos.loop_decide.decide(state, outcome)`; the screenplay's job is to\nconstruct the typed `IterationOutcome` and read the returned `LoopDecision`:\n\n- `action: \"continue\"` → run the next iteration in `next_mode` (`dispatch` or\n  `replan`); if `reconcile` is set (a soft\u002Fdrive STALE-STAMP), run an inline\n  stamp-reconcile pass first. Carry `next_state` forward (the updated counters).\n- `action: \"stop\"` → the loop ends; report `stop_reason` (one of the five above)\n  and `surface` (whether it needs operator attention).\n- `action: \"retry-same-iter\"` → a transient overload; sleep `backoff_seconds`\n  and re-run the SAME iteration.\n\nThe drained-twice rule is the kernel's: a DRAIN counts toward an early stop ONLY\nafter a *productive* `\u002Fdos-replan`. A STALE-STAMP or BLOCKED gate routes to\n`\u002Fdos-replan` (under `hard`) but never arms a false drained-twice stop — that is\nthe structural fix the typed gate exists for.\n\nThe pre-dispatch evidence from Step 1 rides into the decision too: the kernel reads\nthe carried `Pickability` (→ `pick-held-invariant` stop) and `Cooldown` (→\n`pick-cooldown` stop). So the loop's continue\u002Fstop is driven END-TO-END by kernel\nrungs — the honest-STOP that used to be a per-run human override is now a kernel\nrule, not prose the loop re-applies each iteration.\n\n### Step 2b — Gather the outer-ratchet verdict (docs\u002F351, the RSI rung)\n\nOne more piece of in-flight evidence rides into `decide()`: the **outer ratchet**,\nthe verdict on whether this iteration produced a *witnessed net gain* or the loop is\nspinning while narrating progress. The work-metric is the cross-run KEEP the loop\nalready computes in **Step 3** — the count of picks `dos reconcile` confirms VERIFIED\n(exit 0) against git ancestry, NOT the SHIPPED self-report. Carry a cumulative\nbaseline of VERIFIED picks across iterations and ask the kernel:\n\n```bash\n# work = the carried cumulative VERIFIED count + this iteration's VERIFIED picks;\n# baseline_work = the carried cumulative count BEFORE this iteration.\n# suite\u002Ftruth come from the gate being LIVE and reconcile not flagging QUIET_INCOMPLETE.\ndos improve --workspace . \\\n  $( [ \"$GATE_LIVE\" = 1 ] && echo --suite-passed ) \\\n  $( [ \"$RECONCILE_CLEAN\" = 1 ] && echo --truth-clean ) \\\n  --work \"$VERIFIED_TOTAL\" --baseline-work \"$VERIFIED_BASELINE\" \\\n  --consecutive-reverts \"$RATCHET_REVERTS\" --max-reverts 3 --json\n```\n\nBranch on the verdict (the exit code): `0` KEEP — net gain witnessed, raise the\nbaseline (`VERIFIED_BASELINE=$VERIFIED_TOTAL`), reset `RATCHET_REVERTS=0`; `3` REVERT\n— no net gain this iteration, bump `RATCHET_REVERTS`; `4` ESCALATE — the breaker is\nopen. Pass the verdict as `loop_decide.decide`'s `ratchet` evidence. When it is\nESCALATE the kernel STOPs the loop with **not-ratcheting** (condition 8): N iterations\nof \"SHIPPED\" that all reconcile QUIET_INCOMPLETE accrue reverts and stop the loop\ninstead of burning the cap — the exact failure the bare loop hit (motion without\nmeasured progress). This is the same `improve` keep-gate the `\u002Fdos-self-improve` loop\nuses, now gating each dispatch iteration's net gain; the metric is the non-forgeable\nreconcile-VERIFIED count (git ancestry), never the loop's word.\n\n## Step 3 — Reconcile each claimed pick (the cross-run KEEP)\n\nA `dispatch` iteration that SHIPPED claims picks done — but a claim is a\nself-report. Before the archive drops a claimed pick from the residual, reconcile\nits claim against ground truth so a quietly-incomplete pick re-enters the pickable\nset next iteration, flagged:\n\n```bash\ndos reconcile \u003CUNIT> --claimed-done --plan \u003CPLAN> --phase \u003CPHASE>   # oracle from git\n```\n\nBranch on the exit code (the verdict IS the code):\n\n- `0` **VERIFIED** → the oracle confirms it shipped; it leaves the residual.\n- `3` **QUIET_INCOMPLETE** → CLAIMED done but the oracle says NOT_SHIPPED — KEEP it\n  in the residual, flagged; it re-enters the pickable set next iteration so the\n  host routes it (a verifier pass \u002F `\u002Fdos-replan` \u002F a finding). **Do NOT believe\n  the claim** — only ground truth removes work (the FQ-336 touch-counts-as-ship\n  false-DRAIN is exactly what this catches).\n- `4` **HONEST_OPEN** → not claimed, not shipped; honest open work, stays in the\n  residual.\n\nThis is the cross-run KEEP wired at the boundary that runs the write (the\n`CLAUDE.md` \"wire the contract into the step that runs the write\" rule).\n\n## Step 4 — Archive + release (and leave the lane's tree clean)\n\nWhen the loop stops, write a loop record under `paths.runs` (the run dir from\n`dos doctor --json`: the per-iteration verdicts, the reconcile flags, the stop\nreason) and release the lane lease. Commit with a generic subject read from config\n— no hardcoded prefix.\n\n**Leave the tree clean for the lane you held — an unattended loop must not strand\nits own writes.** A loop that ships work but leaves it uncommitted is the bug the\noracle catches next session: `dos verify` answers from git ancestry, so an\nuncommitted change is a phase the kernel reports `NOT_SHIPPED` (the \"a commit IS the\nship-stamp\" contract). So at close-out, commit your lane's writes — driven by the\nsame `dos` verbs the loop already uses, with generic git:\n\n- **Commit your lane's writes by explicit pathspec** — stage and commit exactly the files\n  under the lane region you leased (the `tree` globs `dos lease-lane acquire` granted)\n  via `fak sync check` and `fak commit --path \u003Clane paths> -m \"\u003Csubject>\" [--push]` (or `fak sync push`).\n  (Fallback: `git commit -s -m \"\u003Csubject>\" -- \u003Clane paths>`). **Never a bare `git add -A`** — when sibling loops hold disjoint\n  lanes on the same tree, a blanket add sweeps another loop's in-flight edits into\n  your commit. The lane lease you held names exactly which paths are yours; commit\n  only those.\n- **Carry your worker id as a `(fak-worker \u003Cid>)` trailer (best-effort, #2065).** If\n  `FLEET_WORKER_ID` is set in your environment (the dispatcher stamps it), append a\n  `(fak-worker \u003CFLEET_WORKER_ID>)` trailer line to the commit body so a dispatched\n  worker's witnessed ships are attributable on the shared trunk. This is *additive and\n  best-effort*: it does NOT replace the `(fak \u003Cleaf>)` ship stamp or the `Fixes #N`\n  reference, and it is an attribution **aid, not a witness** — `dispatch_status.py`\n  folds it read-only (`git log --grep='(fak-worker '`) and nothing is gated on it, so a\n  missing trailer never blocks a ship. Skip it silently when `FLEET_WORKER_ID` is unset.\n- **Confirm the tree is clean for your lane before you exit.** `git status\n  --porcelain -- \u003Clane paths>` over the region you leased should come back empty once\n  you have committed. If it does not, you stranded durable work — `log` it and commit\n  it (or, if a path turns out to belong to a *still-live* sibling lease — check\n  `dos lease-lane live`\u002Fthe lane journal — leave it for that loop). Either way the loop\n  must not exit leaving its own lane dirty off a self-reported \"done\".\n- **Scratch is not stranded work.** Short-lived probe output (a host's scratch\n  convention — temp dirs, `*.err`, leading-underscore probes) is deletable noise, not\n  a phase to ship; `rm` it or leave it gitignored, don't commit it into the lane.\n\nThis close-out is what keeps an unattended fleet's tree **clean and well-organized\nacross runs**: each loop commits its own lane and confirms it left nothing behind, so\nthe trunk never accumulates anonymous WIP from a loop that stopped mid-write.\n\n> **DOS-repo note (not part of the generic skill):** when this loop runs *in the DOS\n> kernel repo itself*, that repo ships an advisory `scripts\u002Fgit_hygiene.py --strict`\n> that mechanizes the \"is my lane clean?\" check above (exit 1 on stranded durable\n> work, lease-aware). It is a DOS-repo convenience, not a `dos` verb — a foreign\n> workspace uses the generic `git status` form above. Both express the same\n> discipline.\n\n## What this skill deliberately does NOT do (no silent gap, `CLAUDE.md` heavy tier)\n\n- **No soft-claim lease core.** It coordinates loops by *lane* lease\n  (`dos lease-lane acquire`), not the per-pick soft-claim machinery that stays host-side.\n  `log` this when a sibling lane is busy rather than waiting on a soft-claim.\n- **No value-greedy focus scheduler.** It picks by lane order + the gate verdict,\n  not a host's per-iteration focus ranking. That scheduler is the heavy tier.\n- **No rate-limit predictive monitor \u002F resume manifest.** It stops on a\n  RATE_LIMITED outcome and reports it; it does not pre-empt or auto-resume. A\n  host adds that as a driver concern.\n\nIt `log`s each of these the first time it would have used them, so the capability\ngap is named, never silent.\n\n## Worked example (live transcript)\n\n> **The cadence loop, by hand.** Take a lane, keep the WAL beat alive, ask the\n> kernel if the run is moving. Every line is a real `dos` verb; the heartbeat is\n> the `HEARTBEAT` op the trajectory-audit cross-signal reads.\n\nStep 0 — take a lane. You ask for `src`; a live lease made it contended, so the\narbiter auto-picks a free cluster lane instead (it never double-books a region):\n\n```bash\n$ dos arbitrate --workspace . --lane src\n{\"auto_picked\":true,\"free_clusters\":[],\"lane\":\"benchmark\",\"lane_kind\":\"cluster\",\"outcome\":\"acquire\",\"pick_count\":null,\"reason\":\"auto-picked free cluster lane benchmark (requested src was refused: lane src would edit the orchestrator's own running code … (SELF_MODIFY) …).\",\"tree\":[\"benchmark\u002F**\"]}\n```\n\n→ exit `0` (`acquire`); the redirect IS the admission kernel refusing an\ninadmissible region, with the real reason named in the parenthetical (a free,\nadmissible lane you name is granted directly).\n\nKeep the beat alive across iterations — `acquire` once, `heartbeat` each pass,\n`release` at the end. The `HEARTBEAT` op is what makes `SPINNING` reachable from\nreal evidence (a beat is not an event):\n\n```bash\n$ python tools\u002Fdos_fleet_lease.py --workspace . acquire   --lane benchmark    # writes ACQUIRE, then publishes this node's held set\n$ python tools\u002Fdos_fleet_lease.py --workspace . heartbeat --lane benchmark    # writes HEARTBEAT, then republishes\n$ python tools\u002Fdos_fleet_lease.py --workspace . release   --lane benchmark    # writes RELEASE, then republishes\n```\n\nAsk the kernel if the run is moving (the temporal verdict, from git delta — never\nthe loop's \"making progress\" self-report):\n\n```bash\n$ dos liveness --workspace . --run-id R --start-sha 80d4f30\n```\n\n→ exit `0` = `ADVANCING` · `3` = `SPINNING` · `4` = `STALLED` (`2` contract-error).\nDrive the loop off the exit code, not stdout prose.\n\nRead the beat back from the WAL — an `ACQUIRE`\u002F`HEARTBEAT`\u002F`REFUSE`\u002F`RELEASE`\nsequence on generic lanes (ILLUSTRATIVE shape; `dos journal replay` folds it):\n\n```bash\n$ dos journal --workspace . tail 4\n{\"op\":\"ACQUIRE\",\"lane\":\"benchmark\",\"tree\":[\"benchmark\u002F**\"]}\n{\"op\":\"HEARTBEAT\",\"lane\":\"benchmark\"}\n{\"op\":\"REFUSE\",\"lane\":\"benchmark\",\"reason\":\"lane benchmark is already held by a live loop — pick a different --lane or wait.\"}\n{\"op\":\"RELEASE\",\"lane\":\"benchmark\"}\n```\n\n→ the `REFUSE` row is the WAL recording the arbiter's \"no\" — a sibling loop's\noverlapping lane, fossilized for `dos decisions` to surface.\n\n## Anti-patterns\n\n- ❌ Re-implementing the stop conditions in prose — route every outcome through\n  the kernel loop decision; the counters\u002Fbreakers\u002Fcap are the kernel's.\n- ❌ Counting a STALE-STAMP\u002FBLOCKED toward drained-twice — only a true DRAIN after\n  a productive replan counts (the kernel enforces this; don't second-guess it).\n- ❌ Running two loops on overlapping lanes — the arbiter REFUSEs the second;\n  honor it, don't `--force`.\n","# dos-dispatch-loop — the generic dispatch⇄replan cadence\n\n> **The unattended plan-and-ship loop.** It runs `\u002Fdos-dispatch` repeatedly and\n> falls to `\u002Fdos-replan` when a lane drains, stopping on a typed, kernel-decided\n> condition (not a prose guess). The stop\u002Fcontinue logic is the kernel's\n> `loop_decide.decide` — the loop carries counters, the kernel decides. Several\n> loops on disjoint lanes run in parallel, each holding its own lane lease.\n\nThe stop conditions are the kernel's, in one place:\n\n1. **iteration cap** — reached `max_iterations` (default 10).\n2. **drained-twice** — a DRAIN after a *productive* `\u002Fdos-replan` that itself\n   followed a DRAIN (the lane is genuinely exhausted).\n3. **consecutive-unclear** — the dispatch subprocess is failing systematically.\n4. **rate-limited** — a usage window is exhausted (don't burn launches).\n5. **launch-failed** — a subprocess never started.\n6. **pick-held-invariant** — the next unit is held ONLY by a reason a re-dispatch\n   cannot change (draft-class \u002F operator-gated \u002F soak-open \u002F dependency-unmet);\n   re-dispatching it would re-block identically, so honest-STOP + surface the hold.\n7. **pick-cooldown** — the next unit was attempted-and-didn't-move inside its\n   cooldown window AND nothing fresher is offerable; re-dispatching it would\n   re-storm a known drain (the ~5%-shipping re-pick loop the bare loop hit).\n8. **not-ratcheting** — the OUTER RATCHET (docs\u002F351): the loop has run too many\n   iterations in a row with no *witnessed net gain* — the reconcile-VERIFIED\n   ship-count (Step 3) is not rising, even though each iteration reported SHIPPED.\n   The loop is running but not improving (the \"spinning, narrating progress while\n   net-shipping nothing real\" failure that worsens the longer the loop runs). The\n   verdict is the kernel's `improve` ESCALATE, read VERBATIM — not the loop's\n   self-report. Hand the judgment back to a human.\n\nConditions 6–7 are the docs\u002F207 anti-churn rungs (the loop stops re-picking work\nit cannot move); condition 8 is the docs\u002F351 ratchet (the loop stops when it is\nmoving but not *improving* — RSI made first-class here, the same `improve` keep-gate\nthe `\u002Fdos-self-improve` loop uses, now gating each dispatch iteration's net gain).\n\n## Inputs\n\n- `--lane \u003Cname>` (optional) — focus the whole loop on one lane (fixed for the\n  run; a bare loop auto-picks a free lane at Step 0).\n- `--gate hard|soft|drive` (default `hard`) — the verdict policy. `hard` routes a\n  non-LIVE verdict through `\u002Fdos-replan`; `soft`\u002F`drive` stop on a true DRAIN;\n  `drive` self-heals a STALE-STAMP inline.\n- `--max-iterations \u003CN>` (default 10).\n\n## Step 0 — Pre-flight: take the lane, read the taxonomy\n\n```bash\ndos doctor --workspace . --json\npython tools\u002Fdos_fleet_lease.py --workspace . acquire --lane \u003CLANE> --kind cluster\n```\n\nThe fleet lease wrapper materializes the live lease union, feeds it to the kernel\nacquire, then republishes this node's held set. For ordinary shard lanes this is the\nsame-host fast path; for the three global lanes (`abi`, `release`, `global`) it uses\nthe GitRefStore cross-node transport by default (#21). A REFUSE means a sibling loop\nalready holds an overlapping lane — pick a free one from `kernel.free_clusters` or\nexit. Initialise the loop counters (iteration=1, the breakers at 0).\n\n> **`dos lease-lane acquire` is the lease AUTHORITY; `dos arbitrate` is ADVISORY.**\n> Only `acquire` folds the durable lane-journal WAL under the lease mutex and\n> journals the grant, so only its verdict binds. Honor its REFUSE; never start\n> editing a tree on the strength of an arbitrate GO alone (arbitrate holds no lock,\n> so even a correct GO can be stale by acquire time).\n>\n> The advisory has two surfaces and they do **not** behave the same:\n>\n> - `dos arbitrate` (the **CLI**) reads the live WAL by default, so it agrees with\n>   `acquire`. Passing `--leases '[]'` opts out into an empty world.\n> - The **MCP tool** `dos_arbitrate` reads **nothing** — it arbitrates only against\n>   the `live_leases` you hand it, and an omitted argument means \"the world is\n>   empty\", so every lane reads free. Observed live: MCP `dos_arbitrate` answered\n>   `\"cluster lane 'tools' free — admitted.\"` for a lane `dos lease-lane acquire`\n>   refused the same second as `\"already held by a live loop\"`\n>   ([dos-kernel#246](https:\u002F\u002Fgithub.com\u002Fanthony-chaudhary\u002Fdos-kernel\u002Fissues\u002F246)).\n>   Until that lands, either use the CLI form or pass the real set explicitly\n>   (`dos lease-lane live` \u002F `fak leaseref live`). A FREE from a checker that cannot\n>   see the store is not evidence of a free lane.\n\n## Step 1 — Pick-selection: skip held + cooled units (the anti-churn gate)\n\nBefore a `dispatch` iteration offers a unit, screen the candidate unit set so the\nloop never re-storms work it cannot move (the docs\u002F207 §6 throughline). For each\ncandidate, in order:\n\n```bash\ndos pickable \u003CUNIT> --state '\u003Chost-gathered state>'   # OFFERABLE=0; HELD=per-reason code\ndos cooldown \u003CUNIT>                                    # CLEAR=0; RECENTLY_ATTEMPTED=3\n```\n\n- A `pickable` exit of **0** (OFFERABLE) **and** a `cooldown` exit of **0** (CLEAR)\n  → this unit is dispatchable; offer it and proceed to the iteration.\n- A `pickable` HELD by a re-dispatch-CURABLE reason (IN_FLIGHT \u002F SOFT_CLAIMED \u002F\n  STALE_CLAIM \u002F UNPARSEABLE) or a `cooldown` of **3** (RECENTLY_ATTEMPTED) → **skip\n  this unit, try the next candidate** (the skip-to-next is pick-selection's job).\n- A `pickable` HELD by a re-dispatch-INVARIANT reason (DRAFT_CLASS=10 \u002F\n  OPERATOR_GATED=11 \u002F SOAK_OPEN=12 \u002F DEPENDENCY_UNMET=13) → carry that verdict into\n  Step 2; the kernel will honest-STOP on it (don't re-dispatch a unit a re-dispatch\n  cannot un-gate).\n\nWhen EVERY remaining candidate is skipped (all cooled \u002F curably-held), carry the\nlast `cooldown` RECENTLY_ATTEMPTED (or the invariant `pickable` hold) into Step 2\nas the loop's pre-dispatch evidence — the kernel turns \"nothing fresh is offerable\"\ninto the `pick-cooldown` \u002F `pick-held-invariant` honest-STOP. **Do not re-dispatch\nthe cooled\u002Fheld unit yourself** — that is the re-pick storm this gate prevents.\n\n## Step 1b — Run one iteration\n\nFor a `dispatch` iteration, invoke `\u002Fdos-dispatch --lane \u003CLANE>` (it snapshots,\ngates, and ships). For a `replan` iteration, invoke `\u002Fdos-replan`. Capture the\niteration's outcome:\n\n- a **dispatch** iteration that reached the gate carries a typed verdict — get it\n  from `dos gate` over the packet's dispositions sidecar (LIVE\u002FDRAIN\u002FSTALE-STAMP\u002F\n  BLOCKED\u002FRACE).\n- a **replan** iteration carries a productivity signal (did it refill\u002Fgarden?).\n\n## Step 2 — Decide continue \u002F replan \u002F stop (the kernel decides)\n\nThis is the load-bearing step: **the decision is a kernel mechanism, not prose.**\nFeed the iteration outcome + the carried counters to the loop decider. In code a\nhost calls `dos.loop_decide.decide(state, outcome)`; the screenplay's job is to\nconstruct the typed `IterationOutcome` and read the returned `LoopDecision`:\n\n- `action: \"continue\"` → run the next iteration in `next_mode` (`dispatch` or\n  `replan`); if `reconcile` is set (a soft\u002Fdrive STALE-STAMP), run an inline\n  stamp-reconcile pass first. Carry `next_state` forward (the updated counters).\n- `action: \"stop\"` → the loop ends; report `stop_reason` (one of the five above)\n  and `surface` (whether it needs operator attention).\n- `action: \"retry-same-iter\"` → a transient overload; sleep `backoff_seconds`\n  and re-run the SAME iteration.\n\nThe drained-twice rule is the kernel's: a DRAIN counts toward an early stop ONLY\nafter a *productive* `\u002Fdos-replan`. A STALE-STAMP or BLOCKED gate routes to\n`\u002Fdos-replan` (under `hard`) but never arms a false drained-twice stop — that is\nthe structural fix the typed gate exists for.\n\nThe pre-dispatch evidence from Step 1 rides into the decision too: the kernel reads\nthe carried `Pickability` (→ `pick-held-invariant` stop) and `Cooldown` (→\n`pick-cooldown` stop). So the loop's continue\u002Fstop is driven END-TO-END by kernel\nrungs — the honest-STOP that used to be a per-run human override is now a kernel\nrule, not prose the loop re-applies each iteration.\n\n### Step 2b — Gather the outer-ratchet verdict (docs\u002F351, the RSI rung)\n\nOne more piece of in-flight evidence rides into `decide()`: the **outer ratchet**,\nthe verdict on whether this iteration produced a *witnessed net gain* or the loop is\nspinning while narrating progress. The work-metric is the cross-run KEEP the loop\nalready computes in **Step 3** — the count of picks `dos reconcile` confirms VERIFIED\n(exit 0) against git ancestry, NOT the SHIPPED self-report. Carry a cumulative\nbaseline of VERIFIED picks across iterations and ask the kernel:\n\n```bash\n# work = the carried cumulative VERIFIED count + this iteration's VERIFIED picks;\n# baseline_work = the carried cumulative count BEFORE this iteration.\n# suite\u002Ftruth come from the gate being LIVE and reconcile not flagging QUIET_INCOMPLETE.\ndos improve --workspace . \\\n  $( [ \"$GATE_LIVE\" = 1 ] && echo --suite-passed ) \\\n  $( [ \"$RECONCILE_CLEAN\" = 1 ] && echo --truth-clean ) \\\n  --work \"$VERIFIED_TOTAL\" --baseline-work \"$VERIFIED_BASELINE\" \\\n  --consecutive-reverts \"$RATCHET_REVERTS\" --max-reverts 3 --json\n```\n\nBranch on the verdict (the exit code): `0` KEEP — net gain witnessed, raise the\nbaseline (`VERIFIED_BASELINE=$VERIFIED_TOTAL`), reset `RATCHET_REVERTS=0`; `3` REVERT\n— no net gain this iteration, bump `RATCHET_REVERTS`; `4` ESCALATE — the breaker is\nopen. Pass the verdict as `loop_decide.decide`'s `ratchet` evidence. When it is\nESCALATE the kernel STOPs the loop with **not-ratcheting** (condition 8): N iterations\nof \"SHIPPED\" that all reconcile QUIET_INCOMPLETE accrue reverts and stop the loop\ninstead of burning the cap — the exact failure the bare loop hit (motion without\nmeasured progress). This is the same `improve` keep-gate the `\u002Fdos-self-improve` loop\nuses, now gating each dispatch iteration's net gain; the metric is the non-forgeable\nreconcile-VERIFIED count (git ancestry), never the loop's word.\n\n## Step 3 — Reconcile each claimed pick (the cross-run KEEP)\n\nA `dispatch` iteration that SHIPPED claims picks done — but a claim is a\nself-report. Before the archive drops a claimed pick from the residual, reconcile\nits claim against ground truth so a quietly-incomplete pick re-enters the pickable\nset next iteration, flagged:\n\n```bash\ndos reconcile \u003CUNIT> --claimed-done --plan \u003CPLAN> --phase \u003CPHASE>   # oracle from git\n```\n\nBranch on the exit code (the verdict IS the code):\n\n- `0` **VERIFIED** → the oracle confirms it shipped; it leaves the residual.\n- `3` **QUIET_INCOMPLETE** → CLAIMED done but the oracle says NOT_SHIPPED — KEEP it\n  in the residual, flagged; it re-enters the pickable set next iteration so the\n  host routes it (a verifier pass \u002F `\u002Fdos-replan` \u002F a finding). **Do NOT believe\n  the claim** — only ground truth removes work (the FQ-336 touch-counts-as-ship\n  false-DRAIN is exactly what this catches).\n- `4` **HONEST_OPEN** → not claimed, not shipped; honest open work, stays in the\n  residual.\n\nThis is the cross-run KEEP wired at the boundary that runs the write (the\n`CLAUDE.md` \"wire the contract into the step that runs the write\" rule).\n\n## Step 4 — Archive + release (and leave the lane's tree clean)\n\nWhen the loop stops, write a loop record under `paths.runs` (the run dir from\n`dos doctor --json`: the per-iteration verdicts, the reconcile flags, the stop\nreason) and release the lane lease. Commit with a generic subject read from config\n— no hardcoded prefix.\n\n**Leave the tree clean for the lane you held — an unattended loop must not strand\nits own writes.** A loop that ships work but leaves it uncommitted is the bug the\noracle catches next session: `dos verify` answers from git ancestry, so an\nuncommitted change is a phase the kernel reports `NOT_SHIPPED` (the \"a commit IS the\nship-stamp\" contract). So at close-out, commit your lane's writes — driven by the\nsame `dos` verbs the loop already uses, with generic git:\n\n- **Commit your lane's writes by explicit pathspec** — stage and commit exactly the files\n  under the lane region you leased (the `tree` globs `dos lease-lane acquire` granted)\n  via `fak sync check` and `fak commit --path \u003Clane paths> -m \"\u003Csubject>\" [--push]` (or `fak sync push`).\n  (Fallback: `git commit -s -m \"\u003Csubject>\" -- \u003Clane paths>`). **Never a bare `git add -A`** — when sibling loops hold disjoint\n  lanes on the same tree, a blanket add sweeps another loop's in-flight edits into\n  your commit. The lane lease you held names exactly which paths are yours; commit\n  only those.\n- **Carry your worker id as a `(fak-worker \u003Cid>)` trailer (best-effort, #2065).** If\n  `FLEET_WORKER_ID` is set in your environment (the dispatcher stamps it), append a\n  `(fak-worker \u003CFLEET_WORKER_ID>)` trailer line to the commit body so a dispatched\n  worker's witnessed ships are attributable on the shared trunk. This is *additive and\n  best-effort*: it does NOT replace the `(fak \u003Cleaf>)` ship stamp or the `Fixes #N`\n  reference, and it is an attribution **aid, not a witness** — `dispatch_status.py`\n  folds it read-only (`git log --grep='(fak-worker '`) and nothing is gated on it, so a\n  missing trailer never blocks a ship. Skip it silently when `FLEET_WORKER_ID` is unset.\n- **Confirm the tree is clean for your lane before you exit.** `git status\n  --porcelain -- \u003Clane paths>` over the region you leased should come back empty once\n  you have committed. If it does not, you stranded durable work — `log` it and commit\n  it (or, if a path turns out to belong to a *still-live* sibling lease — check\n  `dos lease-lane live`\u002Fthe lane journal — leave it for that loop). Either way the loop\n  must not exit leaving its own lane dirty off a self-reported \"done\".\n- **Scratch is not stranded work.** Short-lived probe output (a host's scratch\n  convention — temp dirs, `*.err`, leading-underscore probes) is deletable noise, not\n  a phase to ship; `rm` it or leave it gitignored, don't commit it into the lane.\n\nThis close-out is what keeps an unattended fleet's tree **clean and well-organized\nacross runs**: each loop commits its own lane and confirms it left nothing behind, so\nthe trunk never accumulates anonymous WIP from a loop that stopped mid-write.\n\n> **DOS-repo note (not part of the generic skill):** when this loop runs *in the DOS\n> kernel repo itself*, that repo ships an advisory `scripts\u002Fgit_hygiene.py --strict`\n> that mechanizes the \"is my lane clean?\" check above (exit 1 on stranded durable\n> work, lease-aware). It is a DOS-repo convenience, not a `dos` verb — a foreign\n> workspace uses the generic `git status` form above. Both express the same\n> discipline.\n\n## What this skill deliberately does NOT do (no silent gap, `CLAUDE.md` heavy tier)\n\n- **No soft-claim lease core.** It coordinates loops by *lane* lease\n  (`dos lease-lane acquire`), not the per-pick soft-claim machinery that stays host-side.\n  `log` this when a sibling lane is busy rather than waiting on a soft-claim.\n- **No value-greedy focus scheduler.** It picks by lane order + the gate verdict,\n  not a host's per-iteration focus ranking. That scheduler is the heavy tier.\n- **No rate-limit predictive monitor \u002F resume manifest.** It stops on a\n  RATE_LIMITED outcome and reports it; it does not pre-empt or auto-resume. A\n  host adds that as a driver concern.\n\nIt `log`s each of these the first time it would have used them, so the capability\ngap is named, never silent.\n\n## Worked example (live transcript)\n\n> **The cadence loop, by hand.** Take a lane, keep the WAL beat alive, ask the\n> kernel if the run is moving. Every line is a real `dos` verb; the heartbeat is\n> the `HEARTBEAT` op the trajectory-audit cross-signal reads.\n\nStep 0 — take a lane. You ask for `src`; a live lease made it contended, so the\narbiter auto-picks a free cluster lane instead (it never double-books a region):\n\n```bash\n$ dos arbitrate --workspace . --lane src\n{\"auto_picked\":true,\"free_clusters\":[],\"lane\":\"benchmark\",\"lane_kind\":\"cluster\",\"outcome\":\"acquire\",\"pick_count\":null,\"reason\":\"auto-picked free cluster lane benchmark (requested src was refused: lane src would edit the orchestrator's own running code … (SELF_MODIFY) …).\",\"tree\":[\"benchmark\u002F**\"]}\n```\n\n→ exit `0` (`acquire`); the redirect IS the admission kernel refusing an\ninadmissible region, with the real reason named in the parenthetical (a free,\nadmissible lane you name is granted directly).\n\nKeep the beat alive across iterations — `acquire` once, `heartbeat` each pass,\n`release` at the end. The `HEARTBEAT` op is what makes `SPINNING` reachable from\nreal evidence (a beat is not an event):\n\n```bash\n$ python tools\u002Fdos_fleet_lease.py --workspace . acquire   --lane benchmark    # writes ACQUIRE, then publishes this node's held set\n$ python tools\u002Fdos_fleet_lease.py --workspace . heartbeat --lane benchmark    # writes HEARTBEAT, then republishes\n$ python tools\u002Fdos_fleet_lease.py --workspace . release   --lane benchmark    # writes RELEASE, then republishes\n```\n\nAsk the kernel if the run is moving (the temporal verdict, from git delta — never\nthe loop's \"making progress\" self-report):\n\n```bash\n$ dos liveness --workspace . --run-id R --start-sha 80d4f30\n```\n\n→ exit `0` = `ADVANCING` · `3` = `SPINNING` · `4` = `STALLED` (`2` contract-error).\nDrive the loop off the exit code, not stdout prose.\n\nRead the beat back from the WAL — an `ACQUIRE`\u002F`HEARTBEAT`\u002F`REFUSE`\u002F`RELEASE`\nsequence on generic lanes (ILLUSTRATIVE shape; `dos journal replay` folds it):\n\n```bash\n$ dos journal --workspace . tail 4\n{\"op\":\"ACQUIRE\",\"lane\":\"benchmark\",\"tree\":[\"benchmark\u002F**\"]}\n{\"op\":\"HEARTBEAT\",\"lane\":\"benchmark\"}\n{\"op\":\"REFUSE\",\"lane\":\"benchmark\",\"reason\":\"lane benchmark is already held by a live loop — pick a different --lane or wait.\"}\n{\"op\":\"RELEASE\",\"lane\":\"benchmark\"}\n```\n\n→ the `REFUSE` row is the WAL recording the arbiter's \"no\" — a sibling loop's\noverlapping lane, fossilized for `dos decisions` to surface.\n\n## Anti-patterns\n\n- ❌ Re-implementing the stop conditions in prose — route every outcome through\n  the kernel loop decision; the counters\u002Fbreakers\u002Fcap are the kernel's.\n- ❌ Counting a STALE-STAMP\u002FBLOCKED toward drained-twice — only a true DRAIN after\n  a productive replan counts (the kernel enforces this; don't second-guess it).\n- ❌ Running two loops on overlapping lanes — the arbiter REFUSEs the second;\n  honor it, don't `--force`.\n",{"name":5,"description":7},"main","f565b7b7db63e79649f3a02c08108132a86c5d05",".claude\u002Fskills\u002Fdos-dispatch-loop",16,"https:\u002F\u002Fgithub.com\u002Fanthony-chaudhary\u002Ffak","https:\u002F\u002Fanthony-chaudhary.github.io\u002Ffak\u002Fshowcase.html","https:\u002F\u002Fgithub.com\u002Fanthony-chaudhary","https:\u002F\u002Flinkedin.com\u002Fin\u002Fanthonychaudhary","2026-11-15 03:06:00",{"source":8,"skillName":5,"commands":34,"prompt":44,"zipAvailable":45},[35,38,41],{"tool":36,"command":37},"npx","npx skills add anthony-chaudhary\u002Ffak --skill dos-dispatch-loop",{"tool":39,"command":40},"bunx","bunx skills add anthony-chaudhary\u002Ffak --skill dos-dispatch-loop",{"tool":42,"command":43},"pnpm","pnpm dlx skills add anthony-chaudhary\u002Ffak --skill dos-dispatch-loop","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fanthony-chaudhary\u002Ffak\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fdos-dispatch-loop\u002FSKILL.md",true,[],[48,55,64,71,78,83],{"skillNo":49,"name":50,"nameZh":6,"description":51,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":52,"categories":53,"keywords":54,"publishedAt":19},"01M78V4059KRP2T88YR6Z05MJQ","serve-receipt","On-demand research protocol + deterministic receipt for serving a Hugging Face model on a concrete GPU target — fan out research lanes (model facts\u002Ffiles, engine support matrix, serving recipes, KV\u002Fstate memory math), compute the tiered feasibility verdict with `fak serve-receipt` (never narrated math), and emit a witnessed launch receipt. Use when asked for an \"initial config\", \"serving receipt\", or \"how do I serve \u003Cmodel> on \u003CGPUs>\".","https:\u002F\u002Fgithub.com\u002Fanthony-chaudhary\u002Ffak\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fserve-receipt\u002FSKILL.md",[],[],{"skillNo":56,"name":57,"nameZh":6,"description":58,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":59,"licenseFileUrl":60,"sourceHtmlUrl":61,"categories":62,"keywords":63,"publishedAt":19},"01M77HRGKQV2CJ2K6ZFN4YQ32K","caveman","Ultra-compressed communication mode. Cuts output tokens 65% (measured) by speaking like caveman while keeping full technical accuracy. Supports intensity levels: lite, full (default), ultra, wenyan-lite, wenyan-full, wenyan-ultra. Use when user says \"caveman mode\", \"talk like caveman\", \"use caveman\", \"less tokens\", \"be brief\", or invokes \u002Fcaveman. Also auto-triggers when token efficiency is requested.","MIT","https:\u002F\u002Fgithub.com\u002Fanthony-chaudhary\u002Ffak\u002Fblob\u002Fmain\u002Finternal\u002Fsyspromptmmu\u002Fthird_party\u002Fcaveman\u002FLICENSE","https:\u002F\u002Fgithub.com\u002Fanthony-chaudhary\u002Ffak\u002Fblob\u002Fmain\u002Finternal\u002Fsyspromptmmu\u002Fthird_party\u002Fcaveman\u002FSKILL.md",[],[],{"skillNo":65,"name":66,"nameZh":6,"description":67,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":68,"categories":69,"keywords":70,"publishedAt":19},"01M77HRAY3Y3BE7APN5PVJF7NV","quality-report","Score the repo quality and emit a report card.","https:\u002F\u002Fgithub.com\u002Fanthony-chaudhary\u002Ffak\u002Fblob\u002Fmain\u002Finternal\u002Fctxplans\u002Ftestdata\u002Frepo\u002F.claude\u002Fskills\u002Fquality-report\u002FSKILL.md",[],[],{"skillNo":72,"name":73,"nameZh":6,"description":74,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":75,"categories":76,"keywords":77,"publishedAt":19},"01M77HR9807KHAX13DG9PW2K2Q","ctx-overlay","Compact the context window and prune the memory store when a session grows too long.","https:\u002F\u002Fgithub.com\u002Fanthony-chaudhary\u002Ffak\u002Fblob\u002Fmain\u002Finternal\u002Fctxplans\u002Ftestdata\u002Frepo\u002F.claude\u002Fskills\u002Fctx-overlay\u002FSKILL.md",[],[],{"skillNo":79,"name":66,"nameZh":6,"description":67,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":80,"categories":81,"keywords":82,"publishedAt":19},"01M77HR7XQ9X8F2CY9Z4F2ZEZ3","https:\u002F\u002Fgithub.com\u002Fanthony-chaudhary\u002Ffak\u002Fblob\u002Fmain\u002Finternal\u002Fctxplanlint\u002Ftestdata\u002Frepo\u002F.claude\u002Fskills\u002Fquality-report\u002FSKILL.md",[],[],{"skillNo":84,"name":73,"nameZh":6,"description":74,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":85,"categories":86,"keywords":87,"publishedAt":19},"01M77HR5BVKCTR1WNJGRBSDNX7","https:\u002F\u002Fgithub.com\u002Fanthony-chaudhary\u002Ffak\u002Fblob\u002Fmain\u002Finternal\u002Fctxplanlint\u002Ftestdata\u002Frepo\u002F.claude\u002Fskills\u002Fctx-overlay\u002FSKILL.md",[],[]]