[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7DV2W4HZD32VTM3BH5SP7C8":3,"skill-same:01M7DV2W4HZD32VTM3BH5SP7C8":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":6,"ownerDescription":31,"updatedAt":32,"install":33,"resources":46},"01M7DV2W4HZD32VTM3BH5SP7C8","rule-advice-supervision",null,"Contract governing the --advice flag: supervision checkpoints, caps, packets, dossier, receipts, and the boundary that keeps counsel from approving anything. Load only when --advice is present.","ngocsangyem\u002FMeowKit","ngocsangyem","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F33769536?v=4","01M6J3FSMJAG9PVQEGRT7AQ9J2","Sang Nguyen","01M6J3FSRTJG0MNP6NV71639RM",16,"MIT","https:\u002F\u002Fgithub.com\u002Fngocsangyem\u002FMeowKit\u002Fblob\u002Fmain\u002Fkit\u002Fbundles\u002Fcodex\u002Froot\u002F.agents\u002Fskills\u002Frule-advice-supervision\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: \"rule-advice-supervision\"\ndescription: \"Contract governing the --advice flag: supervision checkpoints, caps, packets, dossier, receipts, and the boundary that keeps counsel from approving anything. Load only when --advice is present.\"\n---\n\n# Advice Supervision (`--advice`)\n\nNot a workflow. This is the contract a wrapped skill loads **on demand** when the user\npassed `--advice`, and it is inert otherwise. Do not invoke it as a task.\n\n**Core rule:** when — and only when — the user passes `--advice`, a supporting skill may\ncall the `athena` agent at a small set of **named checkpoints** to get strategic\nsupervision over one delivery run. Athena assesses the situation, makes an evidence-backed\noperational recommendation, guides before work, rescues a blocked run, reviews finished\nwork, and may return work for correction.\n\nSupervision is **evidence a human reads**. It is never an approval, never a gate, never a\nsubstitute for verification.\n\nAthena is a workflow-level strategic intelligence agent with stronger reasoning and\ncross-phase visibility — and **no mutation authority**. The parent owns execution, the\ndossier, receipts and routing. Existing specialists keep their verdicts. Athena is\nreachable through the `--advice` harness, and — where this runtime exposes direct\ndelegation by name — for a stateless strategy consult. It owns no lifecycle phase.\n\nThe mechanical half of this contract is enforced in code by the toolkit CLI — stage and\ndisposition legality, caps, packet size and content checks, dossier fields. Where this\nprose and those checks disagree, the checks are authoritative and the prose is the bug.\n\n## 1 — Activation\n\n- Fires ONLY on an explicit `--advice` flag on a skill that documents support.\n- NEVER from a session-start, session-end, or any other lifecycle hook. Hooks may not\n  invoke agents.\n- NEVER auto-enabled by a mode, tier, risk flag, score, or failure count.\n- Delegated child jobs never self-supervise: a supervised parent does not hand `--advice`\n  to the subagents it spawns.\n\nWithout the flag the checkpoint blocks are inert and cost nothing: the skill does not load\nthis file, makes zero calls, and writes no state or receipt.\n\n## 2 — Cadence\n\n```text\nGUIDE → (executor works) → RESCUE* → REVIEW → RECHECK*\n```\n\nCheckpoints are **macro boundaries**, never per tool call, per file, or per loop\niteration.\n\n| Stage | Purpose | Max per run |\n|---|---|---:|\n| GUIDE | Direction before work starts | 1 |\n| RESCUE | Unblock a stalled or contradicted run | 2 (one per rework round) |\n| REVIEW | Independent read of finished work, against evidence | 1 |\n| RECHECK | Re-examine returned work | 1 |\n\nPer-skill total ceiling:\n\n| Skill | Cap |\n|---|---:|\n| `mk-brainstorming` | 4 |\n| `mk-plan-creator` | 4 |\n| `mk-cook` | 5 |\n| `mk-fix` | 5 |\n| `mk-autobuild` | 5 |\n| `mk-ship` | 4 **per release stage** — `prepare`, `release` and `publish` each carry their own budget, charged by `--release-stage`. The flag is required for ship and refused for every other skill; it is fixed at `begin` and inherited by `commit`. |\n\nRules that keep the cadence bounded:\n\n- A duplicate `checkpointId` is **idempotent** — it returns the prior result and does not\n  consume a cap slot. That is what makes crash-and-resume safe. Ids are unique per RUN,\n  not per partition: reusing one under a different stage or release stage is a naming\n  collision and is refused, because returning another checkpoint's recorded result would\n  silently skip the one the workflow believes it ran.\n- `RECHECK` requires a prior `RETURN_TO_EXECUTOR`.\n- A **second unresolved return escalates to a human.** There is no third opinion. Returns\n  are counted within the budget that spent them, so on a partitioned skill a resolved\n  return in one release stage and an unrelated one in another are two episodes, not one\n  unresolved loop. The escalation flag itself stays run-wide: once a human has been asked,\n  the whole run waits for them.\n- Reaching a cap escalates; it never loops.\n- A recorded call that names no partition, or names one the skill does not declare, counts\n  against **every** partition. Skipping it would turn history written before the skill was\n  partitioned — or a hand-edited value — into a free slot in each budget. Counting it\n  everywhere can only under-permit, which is the direction this contract errs in.\n\n**Rescue complements human stops; it never delays them.** The `mk-fix` rule — 3+ failed\nattempts means stop and question the architecture with the user — fires on its own\nschedule whether or not counsel was taken at two failures. The same holds for the Gate 1\nquestion, the Gate 2 question, and any explicit business decision.\n\n## 3 — Input packet (the parent supplies)\n\nA fresh subagent inherits no conversation, so the packet is the supervisor's entire world.\nFields: `runId`, `skill`, `stage`, `checkpointId`, `mission`, `lockedDecisions`,\n`currentState`, `workerSummary`, `evidenceRefs`, `priorDirective`, `question`,\n`riskAndReversibility`.\n\n- Serialized cap **12 KiB UTF-8** — fail visibly BEFORE delegation, never truncate. A\n  truncated packet asks a different question than the one intended.\n- At most **5 evidence pointers**, each carrying `path`, `relevance`, `provenance` and a\n  short `summary`. Provenance is mandatory: an unattributed path cannot be weighed, and a\n  fresh subagent cannot reconstruct it.\n- **Never** a raw transcript, full diff or log, memory dump, secret or personal data, or\n  unrelated project history. Pass a pointer, not the payload.\n- Start with the map; Athena reads only the detail it selects.\n- Locked decisions and the exact current question appear at both start and end as\n  attention anchors.\n\n## 4 — Output packet (Athena returns)\n\nFields: `disposition`, `strategicAssessment`, `decisionRecommendation`,\n`strategicDirective`, `requiredCorrections`, `nextFalsifiableCheck`, `risksAndRollback`,\n`rejectedAlternatives`, `assumptions`, `confidence`, `evidenceRead`.\n\nDispositions are **stage-specific**, and an illegal one is rejected before routing:\n\n| Stage | Legal dispositions |\n|---|---|\n| GUIDE, RESCUE | `CONTINUE_WITH_DIRECTIVE`, `ESCALATE_TO_HUMAN`, `BLOCKED_MISSING_EVIDENCE` |\n| REVIEW, RECHECK | `READY_FOR_EXISTING_GATE`, `RETURN_TO_EXECUTOR`, `ESCALATE_TO_HUMAN`, `BLOCKED_MISSING_EVIDENCE` |\n\n`READY_FOR_EXISTING_GATE` means \"the normal reviewer or gate is the correct next step\". It\n**never** means the gate is cleared.\n\n- Output cap **600 words**. Volume is not rigor.\n- `requiredCorrections`: max 5, ordered, each independently verifiable and each naming its\n  `proofRequired`. `RETURN_TO_EXECUTOR` requires at least one.\n- `decisionRecommendation` selects one operational path within the packet's locked scope\n  and says why. It MUST escalate rather than override a locked business, security,\n  compliance or gate decision.\n- No `approve`, `clear`, `merge`, `deploy` or equivalent authority language, in any\n  wording. Say \"the evidence supports X\".\n\n### Transport status vs disposition\n\nThe subagent status block is **transport status only**: `DONE` means a valid packet\narrived; `BLOCKED` means no usable directive exists. The `disposition` field is the sole\nworkflow-routing signal. A `BLOCKED` transport may not carry a usable directive, and a\ndelivered packet must state a disposition — either mismatch is refused rather than guessed\nat.\n\n## 5 — Continuity dossier\n\nOne parent-owned file per run at the deterministic path\n`tasks\u002Freports\u002F{supervisionRunId}-athena-supervision.md`.\n\nAthena is a long-lived **lead**, never a long-lived **session**: every call is a fresh\nisolated subagent. Continuity therefore comes from this compact record, NOT from a\ntranscript, an auto-memory, or a new supervisor store.\n\n- Frontmatter and active summary stay under **2 KiB**, holding only run identity, current\n  stage, locked-decision pointers, latest directive, correction count, receipt pointers,\n  and next safe action.\n- Historical receipts may sit below the active summary but are **never auto-loaded**.\n  Never append full model output.\n- The dossier **cannot** carry progress, verification, gate, approval, verdict or status\n  fields. A candidate that does is refused, not stripped — silently dropping a field the\n  caller believed it stored is worse than refusing.\n- Two-step checkpoint marker: `pending` written BEFORE the call, `committed` after the\n  result lands. A pending marker lets a resuming parent recognize an already-attempted\n  checkpoint instead of spending another slot.\n- Pointer placement: an active durable task gets an `evidenceRef`; a plan or run artifact\n  gets a pointer; a one-off run keeps only the local dossier and has no automatic\n  cross-session resume guarantee. Never invent an active task.\n\nA state or write failure disables supervision for that run, emits the exact degraded\nnotice, and lets the ordinary unsupervised workflow continue — but no later Athena call may\nrun until durable state is recovered.\n\n## 6 — Correction cycle\n\n`RETURN_TO_EXECUTOR` routes work back to its **current owner** — planner, developer,\ntester, whoever owns it — never to Athena.\n\n1. The executor addresses each correction or records why it is rejected.\n2. Source changes advance `evidenceRevision` and mark stale verification and review\n   evidence **superseded**; normal checks re-run.\n3. Plan or scope changes additionally invalidate Gate 1 and require a new human approval.\n   In-scope source corrections keep Gate 1 but invalidate downstream evidence.\n4. Athena rechecks once by default. A second unresolved return escalates.\n\nSuperseded evidence must never read as current at a later gate or ship preflight.\n\n**A correction's only write target is an existing `workflow-evidence.json` inside the\nproject.** The index lives under several roots but always under that name, so the name is\nthe check; the CLI refuses any other path. Both the boundary check and the name check run\non the **resolved** path, because a textual check answers a question about the caller's\nstring rather than about the file that will be opened, and reads follow symlinks. A path\nthat does not resolve is absent, and a correction supersedes recorded evidence rather than\ncreating it, so it is refused.\n\n## 6a — Composing with `--html`\n\n`--advice` and `--html` are orthogonal. Neither implies the other, each is parsed\nindependently, and combining them changes no activation, approval, validation, export or\nownership. A run with both produces the same visual artifact, at the same path, validated\nby the same validator, approved by the same human at the same gate, and exported by the\nsame owner as `--html` alone would for the same final Markdown.\n\nThey meet only in the correction loop, because a supervised correction edits Markdown and\nthe visual artifact pins hashes of it. The Markdown plan stays the source of truth; REVIEW\nfires before the gate step where the visual preconditions are checked; and a returned\ncorrection invalidates the artifact's pinned hashes exactly like any other Markdown edit —\nrehash (which clears the prior visual approval), re-validate, re-review, re-approve, then\npresent the ordinary Gate 1. A stale artifact never reaches the gate beside a corrected\nplan.\n\nAthena approves no visual, edits no HTML, changes no filename or path, opens no browser,\nand turns no dossier or receipt into a rendered artifact.\n\n## 7 — Prohibitions\n\nAthena MUST NOT:\n\n- Mutate anything: no source, test, fixture, plan, verdict, status report, or memory\n  write, and no command that alters the working tree. Only the parent writes the dossier\n  and the receipt.\n\n  Be precise about how that is enforced here: a Codex agent definition carries no\n  per-agent tool or permission field, so on this runtime the ban is **behavioral —\n  enforced by the agent's instructions, not by the sandbox**. The other two provider\n  planes enforce it structurally (tool omission, read-only flag). Do not treat the\n  behavioral ban as weaker for that reason, and do not describe this surface as\n  structurally read-only.\n- Interview the user. That is the `advisor` agent behind `mk-advise`, unchanged by this\n  rule.\n- Emit a verdict, score, grade, or security clearance. The `reviewer`, `evaluator` and\n  `security` agents own those.\n- Approve, clear, unblock, or advance Gate 1, Gate 2, a security review, CI, a merge, a\n  deploy, or any user business decision.\n- Change the executor's model, profile, or effort. `--advice` supervises the workflow, not\n  the model running it.\n- Spawn another Athena or any lifecycle agent.\n\n**Gate authority invariant:** automation executes BETWEEN gates and never supplies the\nauthority OF a gate. An Athena directive is in exactly the same class as an evaluator\nverdict or a green test suite — evidence presented to a human at the gate, never the\napproval itself. Prose that lets supervision advance a gate, in any wording, is a\nviolation.\n\nA directive is also **not verification**. Verification comes from tests, review verdicts\nand validators. Supervision never counts toward it.\n\n## 8 — Receipt\n\nAfter each checkpoint the **parent** (never Athena) writes a receipt to\n`tasks\u002Freports\u002F{YYMMDD}-{slug}-advice-{n}.md`:\n\n```markdown\n---\nkind: advice-receipt\nrunId: \u003CsupervisionRunId>\nstage: GUIDE | RESCUE | REVIEW | RECHECK\ndisposition: \u003Cthe returned disposition, verbatim>\noutcome: adopted | rejected | deferred\nreason: \u003Cone line — required even when adopted>\ntaskId: \u003Cactive task id, or \"none\">\nprovider: \u003Cruntime>\nskill: \u003Cskill and workflow>\ncheckpointId: \u003Cnamed checkpoint>\n---\n\nThis is a record of supervision, NEVER verification and never a gate approval.\n\n**Question asked:** …\n**Directive:** … (summary, not the full packet)\n**Required corrections:** … (or \"none\")\n**Evidence pointers:** …\n**Next safe action:** …\n```\n\n`disposition` is Athena's returned routing signal; `outcome` is what the parent did with\nit. They are separate fields because a parent may rightly reject a directive, and\ncollapsing them would hide that. The reason is required even when adopted — \"why\" is what\na later session cannot reconstruct.\n\nA failed receipt write surfaces a one-line notice and never fails silently.\n\n## 9 — Propagation\n\nOnly `supervisionRunId` crosses an approved top-level lifecycle handoff between skills that\nboth document `--advice`. Spawned workers receive a task-specific directive — never the\nflag, never the dossier, never routing ability.\n\n## 10 — Fallback\n\nIf the runtime cannot discover the agent or execute a foreground delegation, the skill\nprints exactly:\n\n```\nadvice checkpoint unavailable in this runtime: \u003Creason>\n```\n\nthen continues unsupervised.\n\n**Inline self-advice impersonating Athena is forbidden.** A recommendation written by the\nagent that is stuck is not an independent check, and labelling it as one corrupts the\nreceipt.\n\n## 11 — Direct consult vs embedded supervision\n\nWhere this runtime exposes direct delegation to the agent by name, that is a **stateless\nstrategy consult**: Athena may assess a difficult situation, compare alternatives and make\na recommended operational decision, but creates no run receipt, correction routing or cap\naccounting. It is NOT lifecycle supervision and must label itself accordingly. It escalates\nrather than changing a locked business, security, compliance or gate decision. Embedded\nsupervision requires a valid `supervisionRunId`.\n\nA direct consult returns a **strategy brief**, which is a different shape from a checkpoint\npacket and deliberately cannot carry a `disposition`:\n\n| Field | Required |\n|---|---|\n| `situation`, `decisionRecommendation`, `nextFalsifiableCheck`, `escalationPoint`, `confidence` | yes |\n| `rejectedAlternatives`, `risksAndRollback`, `assumptions`, `evidenceRead` | no |\n\n`disposition`, `requiredCorrections`, `strategicDirective`, `runId`, `stage`,\n`checkpointId` and any receipt, dossier or correction field are **refused, not stripped**.\nA disposition is a routing signal for a run; emitted without one it reads as a governed\ndecision that no run ever governed, and a later reader cannot tell the difference. That is\nthe impersonation this section exists to prevent, so the refusal lives in a schema rather\nthan in a reviewer's memory.\n\nThe two routes are distinct capabilities, and neither may impersonate the other:\n\n| Route | Requires | Forbidden |\n|---|---|---|\n| `direct` | nothing — it is stateless | any `runId`, `stage` or `checkpointId`; any dossier, receipt or cap accounting |\n| `embedded` | a valid `runId` **and** `stage` **and** `checkpointId` | — |\n\nA call claiming `embedded` without a valid run id is refused rather than downgraded:\nsupervision with nothing to resume from and no cap to bound it is not supervision. A\n`direct` consult that arrives carrying run state is refused rather than stripped, because\nonce that state is written down the consult is indistinguishable from a governed\ncheckpoint. Only `embedded` may produce durable supervision artifacts — structurally, not\nby convention: a dossier path cannot be built without a run id, and a direct consult has\nnone.\n\n## 12 — Model policy\n\n`--advice` never changes the executor's model. Athena maps to the strongest advisory tier\ndeclared in `.codex\u002Fagents\u002Fathena.toml` for this runtime. No model name or id belongs in\nthis contract or in any skill body.\n\n## The call protocol\n\nIdentical at every checkpoint of every wrapped skill, so it lives here once. A wrapper\ndeclares only WHERE its checkpoints fire and WHAT question each asks.\n\n**1. Open the checkpoint.** This is what enforces the cap, stage legality and idempotency,\nand writes the pending marker that makes a crash resumable:\n\n```\nmewkit advice begin --run \u003CsupervisionRunId> --skill mk-\u003Cskill> \\\n  --stage GUIDE|RESCUE|REVIEW|RECHECK --checkpoint \u003CcheckpointId>\n```\n\n`mk-ship` additionally passes `--release-stage prepare|release|publish`; every other skill\nis refused if it passes one. A refusal is final for that checkpoint: continue unsupervised,\nor escalate when the refusal says to. Re-running the same `--checkpoint` returns the\nrecorded result and spends no slot.\n\n**2. Validate the packet, then delegate.** The caps, pointer budget, provenance requirement\nand secret scan are enforced by the command, not by writing the packet carefully:\n\n```\nmewkit advice validate-packet --evidence \u003Cpacket.json> --packet-kind input\n```\n\nThen delegate to the `athena` agent by name, with the packet inline — a fresh subagent\ninherits no conversation. Validate the returned packet the same way with\n`--packet-kind output` before acting on it or summarizing it into a receipt.\n\n**3. Commit the result.** Render the returned packet to the user, then:\n\n```\nmewkit advice commit --run \u003CrunId> --checkpoint \u003CcheckpointId> \\\n  --disposition \u003Creturned disposition> --outcome adopted|rejected|deferred \\\n  --reason \"\u003Cone line, required even when adopted>\" \\\n  --directive \"\u003Csummary>\" --next \"\u003Cnext safe action>\" \\\n  [--correction \"\u003Cchange>\" ...] [--evidence-pointer \u003Cpath> ...]\n```\n\n`commit` writes the receipt, records the call against the cap, and refuses a disposition\nthat is illegal for the stage. `disposition` is Athena's routing signal; `--outcome` is\nwhat the wrapper decided to do with it — a rejected directive is a legitimate, recordable\noutcome.\n\n**4. Route on the disposition.**\n\n- `CONTINUE_WITH_DIRECTIVE` — proceed; the directive is input, not instruction.\n- `READY_FOR_EXISTING_GATE` — run the skill's normal next gate or review. It is NOT cleared.\n- `RETURN_TO_EXECUTOR` — apply the corrections, then supersede the stale evidence:\n  `mewkit advice commit … --disposition RETURN_TO_EXECUTOR --evidence \u003Cworkflow-evidence.json> --correction-kind source|scope`.\n  Re-run the skill's verification before its review. `scope` also returns Gate 1 to\n  `required`.\n- `ESCALATE_TO_HUMAN` — stop at the existing human touchpoint.\n- `BLOCKED_MISSING_EVIDENCE` — supply the named evidence or continue unsupervised.\n\nIf an active durable task record exists, point at the receipt with\n`mewkit task-state update \u003Cid> --evidence-ref \u003Creceipt path>`. With no active record, keep\nthe file and skip this step — never invent a record.\n\n## Wiring status on this runtime\n\nThe adapter is authored: `.codex\u002Fagents\u002Fathena.toml` carries the full supervisor contract.\n\nSix wrappers declare the flag and name their checkpoints: `mk-brainstorming` (deep\nworkflow only), `mk-plan-creator`, `mk-cook`, `mk-fix`, `mk-autobuild` and `mk-ship`. Each\nfires at the stage boundaries in §2 and drives `mewkit advice begin|commit`, which is where\nthe caps, stage legality, idempotency, dossier and receipt are actually enforced.\n\n`mk-workflow-orchestrator` exposes **no** flag and is not an entry point. When an\nexplicitly enabled run hands off, it carries `supervisionRunId` forward as an opaque value\nat macro boundaries and nothing else.\n\nEvery other surface is excluded, and the exclusion is enforced in code rather than by\nconvention: `mewkit advice begin` refuses a skill absent from the cap table.\n\n**Delegation on this runtime is authored but not yet live-verified.** No session has\nexercised reaching the agent from a wrapper here. If the runtime cannot discover or\ndelegate to `athena`, print the fallback line in §10 and continue unsupervised. Never\nimprovise a checkpoint: writing a packet inline and presenting it as the agent's output is\nnot an independent check, and a receipt produced that way records supervision that never\nhappened.\n\n### Parent-side commands\n\n| Command | Enforces |\n|---|---|\n| `mewkit advice begin` | route contract, supervised-skill check, per-stage and per-skill caps, partition legality (`--release-stage`), idempotent `checkpointId`, pending marker |\n| `mewkit advice commit` | stage and disposition legality, receipt validation (authority language, credentials, empty return), dossier commit, correction supersession |\n| `mewkit advice status` | resume view: current stage, calls used, latest directive, next safe action |\n| `mewkit advice validate-packet` | input packet, output packet, or direct-consult brief — `--packet-kind input\\|output\\|brief` |\n\nA corrupt dossier is refused, never read as a fresh run — otherwise breaking the file that\ncounts the calls would be the cheapest way to buy unlimited ones.\n\n## Gotchas\n\n- Loading this file when `--advice` was not passed — it is inert then, and the load is pure\n  context cost.\n- Treating the behavioral write-ban as weaker because no sandbox enforces it on this\n  runtime.\n- Using a GUIDE disposition at REVIEW, or vice versa — the stage decides what is legal.\n- Emitting a `disposition` from a direct consult — that shape is a checkpoint result and a\n  consult has no run to govern.\n- Reading `READY_FOR_EXISTING_GATE` as a cleared gate — it names the next step, nothing\n  more.\n","# Advice Supervision (`--advice`)\n\nNot a workflow. This is the contract a wrapped skill loads **on demand** when the user\npassed `--advice`, and it is inert otherwise. Do not invoke it as a task.\n\n**Core rule:** when — and only when — the user passes `--advice`, a supporting skill may\ncall the `athena` agent at a small set of **named checkpoints** to get strategic\nsupervision over one delivery run. Athena assesses the situation, makes an evidence-backed\noperational recommendation, guides before work, rescues a blocked run, reviews finished\nwork, and may return work for correction.\n\nSupervision is **evidence a human reads**. It is never an approval, never a gate, never a\nsubstitute for verification.\n\nAthena is a workflow-level strategic intelligence agent with stronger reasoning and\ncross-phase visibility — and **no mutation authority**. The parent owns execution, the\ndossier, receipts and routing. Existing specialists keep their verdicts. Athena is\nreachable through the `--advice` harness, and — where this runtime exposes direct\ndelegation by name — for a stateless strategy consult. It owns no lifecycle phase.\n\nThe mechanical half of this contract is enforced in code by the toolkit CLI — stage and\ndisposition legality, caps, packet size and content checks, dossier fields. Where this\nprose and those checks disagree, the checks are authoritative and the prose is the bug.\n\n## 1 — Activation\n\n- Fires ONLY on an explicit `--advice` flag on a skill that documents support.\n- NEVER from a session-start, session-end, or any other lifecycle hook. Hooks may not\n  invoke agents.\n- NEVER auto-enabled by a mode, tier, risk flag, score, or failure count.\n- Delegated child jobs never self-supervise: a supervised parent does not hand `--advice`\n  to the subagents it spawns.\n\nWithout the flag the checkpoint blocks are inert and cost nothing: the skill does not load\nthis file, makes zero calls, and writes no state or receipt.\n\n## 2 — Cadence\n\n```text\nGUIDE → (executor works) → RESCUE* → REVIEW → RECHECK*\n```\n\nCheckpoints are **macro boundaries**, never per tool call, per file, or per loop\niteration.\n\n| Stage | Purpose | Max per run |\n|---|---|---:|\n| GUIDE | Direction before work starts | 1 |\n| RESCUE | Unblock a stalled or contradicted run | 2 (one per rework round) |\n| REVIEW | Independent read of finished work, against evidence | 1 |\n| RECHECK | Re-examine returned work | 1 |\n\nPer-skill total ceiling:\n\n| Skill | Cap |\n|---|---:|\n| `mk-brainstorming` | 4 |\n| `mk-plan-creator` | 4 |\n| `mk-cook` | 5 |\n| `mk-fix` | 5 |\n| `mk-autobuild` | 5 |\n| `mk-ship` | 4 **per release stage** — `prepare`, `release` and `publish` each carry their own budget, charged by `--release-stage`. The flag is required for ship and refused for every other skill; it is fixed at `begin` and inherited by `commit`. |\n\nRules that keep the cadence bounded:\n\n- A duplicate `checkpointId` is **idempotent** — it returns the prior result and does not\n  consume a cap slot. That is what makes crash-and-resume safe. Ids are unique per RUN,\n  not per partition: reusing one under a different stage or release stage is a naming\n  collision and is refused, because returning another checkpoint's recorded result would\n  silently skip the one the workflow believes it ran.\n- `RECHECK` requires a prior `RETURN_TO_EXECUTOR`.\n- A **second unresolved return escalates to a human.** There is no third opinion. Returns\n  are counted within the budget that spent them, so on a partitioned skill a resolved\n  return in one release stage and an unrelated one in another are two episodes, not one\n  unresolved loop. The escalation flag itself stays run-wide: once a human has been asked,\n  the whole run waits for them.\n- Reaching a cap escalates; it never loops.\n- A recorded call that names no partition, or names one the skill does not declare, counts\n  against **every** partition. Skipping it would turn history written before the skill was\n  partitioned — or a hand-edited value — into a free slot in each budget. Counting it\n  everywhere can only under-permit, which is the direction this contract errs in.\n\n**Rescue complements human stops; it never delays them.** The `mk-fix` rule — 3+ failed\nattempts means stop and question the architecture with the user — fires on its own\nschedule whether or not counsel was taken at two failures. The same holds for the Gate 1\nquestion, the Gate 2 question, and any explicit business decision.\n\n## 3 — Input packet (the parent supplies)\n\nA fresh subagent inherits no conversation, so the packet is the supervisor's entire world.\nFields: `runId`, `skill`, `stage`, `checkpointId`, `mission`, `lockedDecisions`,\n`currentState`, `workerSummary`, `evidenceRefs`, `priorDirective`, `question`,\n`riskAndReversibility`.\n\n- Serialized cap **12 KiB UTF-8** — fail visibly BEFORE delegation, never truncate. A\n  truncated packet asks a different question than the one intended.\n- At most **5 evidence pointers**, each carrying `path`, `relevance`, `provenance` and a\n  short `summary`. Provenance is mandatory: an unattributed path cannot be weighed, and a\n  fresh subagent cannot reconstruct it.\n- **Never** a raw transcript, full diff or log, memory dump, secret or personal data, or\n  unrelated project history. Pass a pointer, not the payload.\n- Start with the map; Athena reads only the detail it selects.\n- Locked decisions and the exact current question appear at both start and end as\n  attention anchors.\n\n## 4 — Output packet (Athena returns)\n\nFields: `disposition`, `strategicAssessment`, `decisionRecommendation`,\n`strategicDirective`, `requiredCorrections`, `nextFalsifiableCheck`, `risksAndRollback`,\n`rejectedAlternatives`, `assumptions`, `confidence`, `evidenceRead`.\n\nDispositions are **stage-specific**, and an illegal one is rejected before routing:\n\n| Stage | Legal dispositions |\n|---|---|\n| GUIDE, RESCUE | `CONTINUE_WITH_DIRECTIVE`, `ESCALATE_TO_HUMAN`, `BLOCKED_MISSING_EVIDENCE` |\n| REVIEW, RECHECK | `READY_FOR_EXISTING_GATE`, `RETURN_TO_EXECUTOR`, `ESCALATE_TO_HUMAN`, `BLOCKED_MISSING_EVIDENCE` |\n\n`READY_FOR_EXISTING_GATE` means \"the normal reviewer or gate is the correct next step\". It\n**never** means the gate is cleared.\n\n- Output cap **600 words**. Volume is not rigor.\n- `requiredCorrections`: max 5, ordered, each independently verifiable and each naming its\n  `proofRequired`. `RETURN_TO_EXECUTOR` requires at least one.\n- `decisionRecommendation` selects one operational path within the packet's locked scope\n  and says why. It MUST escalate rather than override a locked business, security,\n  compliance or gate decision.\n- No `approve`, `clear`, `merge`, `deploy` or equivalent authority language, in any\n  wording. Say \"the evidence supports X\".\n\n### Transport status vs disposition\n\nThe subagent status block is **transport status only**: `DONE` means a valid packet\narrived; `BLOCKED` means no usable directive exists. The `disposition` field is the sole\nworkflow-routing signal. A `BLOCKED` transport may not carry a usable directive, and a\ndelivered packet must state a disposition — either mismatch is refused rather than guessed\nat.\n\n## 5 — Continuity dossier\n\nOne parent-owned file per run at the deterministic path\n`tasks\u002Freports\u002F{supervisionRunId}-athena-supervision.md`.\n\nAthena is a long-lived **lead**, never a long-lived **session**: every call is a fresh\nisolated subagent. Continuity therefore comes from this compact record, NOT from a\ntranscript, an auto-memory, or a new supervisor store.\n\n- Frontmatter and active summary stay under **2 KiB**, holding only run identity, current\n  stage, locked-decision pointers, latest directive, correction count, receipt pointers,\n  and next safe action.\n- Historical receipts may sit below the active summary but are **never auto-loaded**.\n  Never append full model output.\n- The dossier **cannot** carry progress, verification, gate, approval, verdict or status\n  fields. A candidate that does is refused, not stripped — silently dropping a field the\n  caller believed it stored is worse than refusing.\n- Two-step checkpoint marker: `pending` written BEFORE the call, `committed` after the\n  result lands. A pending marker lets a resuming parent recognize an already-attempted\n  checkpoint instead of spending another slot.\n- Pointer placement: an active durable task gets an `evidenceRef`; a plan or run artifact\n  gets a pointer; a one-off run keeps only the local dossier and has no automatic\n  cross-session resume guarantee. Never invent an active task.\n\nA state or write failure disables supervision for that run, emits the exact degraded\nnotice, and lets the ordinary unsupervised workflow continue — but no later Athena call may\nrun until durable state is recovered.\n\n## 6 — Correction cycle\n\n`RETURN_TO_EXECUTOR` routes work back to its **current owner** — planner, developer,\ntester, whoever owns it — never to Athena.\n\n1. The executor addresses each correction or records why it is rejected.\n2. Source changes advance `evidenceRevision` and mark stale verification and review\n   evidence **superseded**; normal checks re-run.\n3. Plan or scope changes additionally invalidate Gate 1 and require a new human approval.\n   In-scope source corrections keep Gate 1 but invalidate downstream evidence.\n4. Athena rechecks once by default. A second unresolved return escalates.\n\nSuperseded evidence must never read as current at a later gate or ship preflight.\n\n**A correction's only write target is an existing `workflow-evidence.json` inside the\nproject.** The index lives under several roots but always under that name, so the name is\nthe check; the CLI refuses any other path. Both the boundary check and the name check run\non the **resolved** path, because a textual check answers a question about the caller's\nstring rather than about the file that will be opened, and reads follow symlinks. A path\nthat does not resolve is absent, and a correction supersedes recorded evidence rather than\ncreating it, so it is refused.\n\n## 6a — Composing with `--html`\n\n`--advice` and `--html` are orthogonal. Neither implies the other, each is parsed\nindependently, and combining them changes no activation, approval, validation, export or\nownership. A run with both produces the same visual artifact, at the same path, validated\nby the same validator, approved by the same human at the same gate, and exported by the\nsame owner as `--html` alone would for the same final Markdown.\n\nThey meet only in the correction loop, because a supervised correction edits Markdown and\nthe visual artifact pins hashes of it. The Markdown plan stays the source of truth; REVIEW\nfires before the gate step where the visual preconditions are checked; and a returned\ncorrection invalidates the artifact's pinned hashes exactly like any other Markdown edit —\nrehash (which clears the prior visual approval), re-validate, re-review, re-approve, then\npresent the ordinary Gate 1. A stale artifact never reaches the gate beside a corrected\nplan.\n\nAthena approves no visual, edits no HTML, changes no filename or path, opens no browser,\nand turns no dossier or receipt into a rendered artifact.\n\n## 7 — Prohibitions\n\nAthena MUST NOT:\n\n- Mutate anything: no source, test, fixture, plan, verdict, status report, or memory\n  write, and no command that alters the working tree. Only the parent writes the dossier\n  and the receipt.\n\n  Be precise about how that is enforced here: a Codex agent definition carries no\n  per-agent tool or permission field, so on this runtime the ban is **behavioral —\n  enforced by the agent's instructions, not by the sandbox**. The other two provider\n  planes enforce it structurally (tool omission, read-only flag). Do not treat the\n  behavioral ban as weaker for that reason, and do not describe this surface as\n  structurally read-only.\n- Interview the user. That is the `advisor` agent behind `mk-advise`, unchanged by this\n  rule.\n- Emit a verdict, score, grade, or security clearance. The `reviewer`, `evaluator` and\n  `security` agents own those.\n- Approve, clear, unblock, or advance Gate 1, Gate 2, a security review, CI, a merge, a\n  deploy, or any user business decision.\n- Change the executor's model, profile, or effort. `--advice` supervises the workflow, not\n  the model running it.\n- Spawn another Athena or any lifecycle agent.\n\n**Gate authority invariant:** automation executes BETWEEN gates and never supplies the\nauthority OF a gate. An Athena directive is in exactly the same class as an evaluator\nverdict or a green test suite — evidence presented to a human at the gate, never the\napproval itself. Prose that lets supervision advance a gate, in any wording, is a\nviolation.\n\nA directive is also **not verification**. Verification comes from tests, review verdicts\nand validators. Supervision never counts toward it.\n\n## 8 — Receipt\n\nAfter each checkpoint the **parent** (never Athena) writes a receipt to\n`tasks\u002Freports\u002F{YYMMDD}-{slug}-advice-{n}.md`:\n\n```markdown\n---\nkind: advice-receipt\nrunId: \u003CsupervisionRunId>\nstage: GUIDE | RESCUE | REVIEW | RECHECK\ndisposition: \u003Cthe returned disposition, verbatim>\noutcome: adopted | rejected | deferred\nreason: \u003Cone line — required even when adopted>\ntaskId: \u003Cactive task id, or \"none\">\nprovider: \u003Cruntime>\nskill: \u003Cskill and workflow>\ncheckpointId: \u003Cnamed checkpoint>\n---\n\nThis is a record of supervision, NEVER verification and never a gate approval.\n\n**Question asked:** …\n**Directive:** … (summary, not the full packet)\n**Required corrections:** … (or \"none\")\n**Evidence pointers:** …\n**Next safe action:** …\n```\n\n`disposition` is Athena's returned routing signal; `outcome` is what the parent did with\nit. They are separate fields because a parent may rightly reject a directive, and\ncollapsing them would hide that. The reason is required even when adopted — \"why\" is what\na later session cannot reconstruct.\n\nA failed receipt write surfaces a one-line notice and never fails silently.\n\n## 9 — Propagation\n\nOnly `supervisionRunId` crosses an approved top-level lifecycle handoff between skills that\nboth document `--advice`. Spawned workers receive a task-specific directive — never the\nflag, never the dossier, never routing ability.\n\n## 10 — Fallback\n\nIf the runtime cannot discover the agent or execute a foreground delegation, the skill\nprints exactly:\n\n```\nadvice checkpoint unavailable in this runtime: \u003Creason>\n```\n\nthen continues unsupervised.\n\n**Inline self-advice impersonating Athena is forbidden.** A recommendation written by the\nagent that is stuck is not an independent check, and labelling it as one corrupts the\nreceipt.\n\n## 11 — Direct consult vs embedded supervision\n\nWhere this runtime exposes direct delegation to the agent by name, that is a **stateless\nstrategy consult**: Athena may assess a difficult situation, compare alternatives and make\na recommended operational decision, but creates no run receipt, correction routing or cap\naccounting. It is NOT lifecycle supervision and must label itself accordingly. It escalates\nrather than changing a locked business, security, compliance or gate decision. Embedded\nsupervision requires a valid `supervisionRunId`.\n\nA direct consult returns a **strategy brief**, which is a different shape from a checkpoint\npacket and deliberately cannot carry a `disposition`:\n\n| Field | Required |\n|---|---|\n| `situation`, `decisionRecommendation`, `nextFalsifiableCheck`, `escalationPoint`, `confidence` | yes |\n| `rejectedAlternatives`, `risksAndRollback`, `assumptions`, `evidenceRead` | no |\n\n`disposition`, `requiredCorrections`, `strategicDirective`, `runId`, `stage`,\n`checkpointId` and any receipt, dossier or correction field are **refused, not stripped**.\nA disposition is a routing signal for a run; emitted without one it reads as a governed\ndecision that no run ever governed, and a later reader cannot tell the difference. That is\nthe impersonation this section exists to prevent, so the refusal lives in a schema rather\nthan in a reviewer's memory.\n\nThe two routes are distinct capabilities, and neither may impersonate the other:\n\n| Route | Requires | Forbidden |\n|---|---|---|\n| `direct` | nothing — it is stateless | any `runId`, `stage` or `checkpointId`; any dossier, receipt or cap accounting |\n| `embedded` | a valid `runId` **and** `stage` **and** `checkpointId` | — |\n\nA call claiming `embedded` without a valid run id is refused rather than downgraded:\nsupervision with nothing to resume from and no cap to bound it is not supervision. A\n`direct` consult that arrives carrying run state is refused rather than stripped, because\nonce that state is written down the consult is indistinguishable from a governed\ncheckpoint. Only `embedded` may produce durable supervision artifacts — structurally, not\nby convention: a dossier path cannot be built without a run id, and a direct consult has\nnone.\n\n## 12 — Model policy\n\n`--advice` never changes the executor's model. Athena maps to the strongest advisory tier\ndeclared in `.codex\u002Fagents\u002Fathena.toml` for this runtime. No model name or id belongs in\nthis contract or in any skill body.\n\n## The call protocol\n\nIdentical at every checkpoint of every wrapped skill, so it lives here once. A wrapper\ndeclares only WHERE its checkpoints fire and WHAT question each asks.\n\n**1. Open the checkpoint.** This is what enforces the cap, stage legality and idempotency,\nand writes the pending marker that makes a crash resumable:\n\n```\nmewkit advice begin --run \u003CsupervisionRunId> --skill mk-\u003Cskill> \\\n  --stage GUIDE|RESCUE|REVIEW|RECHECK --checkpoint \u003CcheckpointId>\n```\n\n`mk-ship` additionally passes `--release-stage prepare|release|publish`; every other skill\nis refused if it passes one. A refusal is final for that checkpoint: continue unsupervised,\nor escalate when the refusal says to. Re-running the same `--checkpoint` returns the\nrecorded result and spends no slot.\n\n**2. Validate the packet, then delegate.** The caps, pointer budget, provenance requirement\nand secret scan are enforced by the command, not by writing the packet carefully:\n\n```\nmewkit advice validate-packet --evidence \u003Cpacket.json> --packet-kind input\n```\n\nThen delegate to the `athena` agent by name, with the packet inline — a fresh subagent\ninherits no conversation. Validate the returned packet the same way with\n`--packet-kind output` before acting on it or summarizing it into a receipt.\n\n**3. Commit the result.** Render the returned packet to the user, then:\n\n```\nmewkit advice commit --run \u003CrunId> --checkpoint \u003CcheckpointId> \\\n  --disposition \u003Creturned disposition> --outcome adopted|rejected|deferred \\\n  --reason \"\u003Cone line, required even when adopted>\" \\\n  --directive \"\u003Csummary>\" --next \"\u003Cnext safe action>\" \\\n  [--correction \"\u003Cchange>\" ...] [--evidence-pointer \u003Cpath> ...]\n```\n\n`commit` writes the receipt, records the call against the cap, and refuses a disposition\nthat is illegal for the stage. `disposition` is Athena's routing signal; `--outcome` is\nwhat the wrapper decided to do with it — a rejected directive is a legitimate, recordable\noutcome.\n\n**4. Route on the disposition.**\n\n- `CONTINUE_WITH_DIRECTIVE` — proceed; the directive is input, not instruction.\n- `READY_FOR_EXISTING_GATE` — run the skill's normal next gate or review. It is NOT cleared.\n- `RETURN_TO_EXECUTOR` — apply the corrections, then supersede the stale evidence:\n  `mewkit advice commit … --disposition RETURN_TO_EXECUTOR --evidence \u003Cworkflow-evidence.json> --correction-kind source|scope`.\n  Re-run the skill's verification before its review. `scope` also returns Gate 1 to\n  `required`.\n- `ESCALATE_TO_HUMAN` — stop at the existing human touchpoint.\n- `BLOCKED_MISSING_EVIDENCE` — supply the named evidence or continue unsupervised.\n\nIf an active durable task record exists, point at the receipt with\n`mewkit task-state update \u003Cid> --evidence-ref \u003Creceipt path>`. With no active record, keep\nthe file and skip this step — never invent a record.\n\n## Wiring status on this runtime\n\nThe adapter is authored: `.codex\u002Fagents\u002Fathena.toml` carries the full supervisor contract.\n\nSix wrappers declare the flag and name their checkpoints: `mk-brainstorming` (deep\nworkflow only), `mk-plan-creator`, `mk-cook`, `mk-fix`, `mk-autobuild` and `mk-ship`. Each\nfires at the stage boundaries in §2 and drives `mewkit advice begin|commit`, which is where\nthe caps, stage legality, idempotency, dossier and receipt are actually enforced.\n\n`mk-workflow-orchestrator` exposes **no** flag and is not an entry point. When an\nexplicitly enabled run hands off, it carries `supervisionRunId` forward as an opaque value\nat macro boundaries and nothing else.\n\nEvery other surface is excluded, and the exclusion is enforced in code rather than by\nconvention: `mewkit advice begin` refuses a skill absent from the cap table.\n\n**Delegation on this runtime is authored but not yet live-verified.** No session has\nexercised reaching the agent from a wrapper here. If the runtime cannot discover or\ndelegate to `athena`, print the fallback line in §10 and continue unsupervised. Never\nimprovise a checkpoint: writing a packet inline and presenting it as the agent's output is\nnot an independent check, and a receipt produced that way records supervision that never\nhappened.\n\n### Parent-side commands\n\n| Command | Enforces |\n|---|---|\n| `mewkit advice begin` | route contract, supervised-skill check, per-stage and per-skill caps, partition legality (`--release-stage`), idempotent `checkpointId`, pending marker |\n| `mewkit advice commit` | stage and disposition legality, receipt validation (authority language, credentials, empty return), dossier commit, correction supersession |\n| `mewkit advice status` | resume view: current stage, calls used, latest directive, next safe action |\n| `mewkit advice validate-packet` | input packet, output packet, or direct-consult brief — `--packet-kind input\\|output\\|brief` |\n\nA corrupt dossier is refused, never read as a fresh run — otherwise breaking the file that\ncounts the calls would be the cheapest way to buy unlimited ones.\n\n## Gotchas\n\n- Loading this file when `--advice` was not passed — it is inert then, and the load is pure\n  context cost.\n- Treating the behavioral write-ban as weaker because no sandbox enforces it on this\n  runtime.\n- Using a GUIDE disposition at REVIEW, or vice versa — the stage decides what is legal.\n- Emitting a `disposition` from a direct consult — that shape is a checkpoint result and a\n  consult has no run to govern.\n- Reading `READY_FOR_EXISTING_GATE` as a cleared gate — it names the next step, nothing\n  more.\n",{"name":5,"description":7},"main","c7ff0a9e672de2c0590b03f6459385c2d7a63c1a","kit\u002Fbundles\u002Fcodex\u002Froot\u002F.agents\u002Fskills\u002Frule-advice-supervision",3,"https:\u002F\u002Fgithub.com\u002Fngocsangyem\u002FMeowKit","https:\u002F\u002Fdocs.meowkit.dev\u002F","https:\u002F\u002Fgithub.com\u002Fngocsangyem","Keep moving forward","2026-11-15 03:54:21",{"source":8,"skillName":5,"commands":34,"prompt":44,"zipAvailable":45},[35,38,41],{"tool":36,"command":37},"npx","npx skills add ngocsangyem\u002FMeowKit --skill rule-advice-supervision",{"tool":39,"command":40},"bunx","bunx skills add ngocsangyem\u002FMeowKit --skill rule-advice-supervision",{"tool":42,"command":43},"pnpm","pnpm dlx skills add ngocsangyem\u002FMeowKit --skill rule-advice-supervision","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fngocsangyem\u002FMeowKit\u002Fblob\u002Fmain\u002Fkit\u002Fbundles\u002Fcodex\u002Froot\u002F.agents\u002Fskills\u002Frule-advice-supervision\u002FSKILL.md",true,[],[48,55,62,69,76,83],{"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},"01M7DVHF481B4J41T4855KGM50","mk-worktree","Manages git worktrees for parallel agent isolation: creates parallel\u002F{name}-{timestamp} branches, merges with conflict detection. NOT for user feature branches (use git worktree directly).","https:\u002F\u002Fgithub.com\u002Fngocsangyem\u002FMeowKit\u002Fblob\u002Fmain\u002Fkit\u002Fbundles\u002Fcursor\u002Froot\u002F.cursor\u002Fskills\u002Fmk-worktree\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":15,"licenseFileUrl":6,"sourceHtmlUrl":59,"categories":60,"keywords":61,"publishedAt":19},"01M7DVHAWVKDKPB1EA3F1NC4JC","mk-workflow-orchestrator","Auto-invoked 7-phase workflow for complex-feature intent; fasttrack mode for pre-approved specs. NOT for explicit single tasks (mk:cook), green-field builds (mk:autobuild), or simple fixes (mk:fix).","https:\u002F\u002Fgithub.com\u002Fngocsangyem\u002FMeowKit\u002Fblob\u002Fmain\u002Fkit\u002Fbundles\u002Fcursor\u002Froot\u002F.cursor\u002Fskills\u002Fmk-workflow-orchestrator\u002FSKILL.md",[],[],{"skillNo":63,"name":64,"nameZh":6,"description":65,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":66,"categories":67,"keywords":68,"publishedAt":19},"01M7DVH6SYBHWW00TRZXCC56JW","mk-web-to-markdown","Fetch a URL and return clean markdown, for arbitrary external pages (blog, RFC, GitHub issue, vendor doc) not covered by mk:docs-finder. Static-only by default; opt-in JS rendering available.","https:\u002F\u002Fgithub.com\u002Fngocsangyem\u002FMeowKit\u002Fblob\u002Fmain\u002Fkit\u002Fbundles\u002Fcursor\u002Froot\u002F.cursor\u002Fskills\u002Fmk-web-to-markdown\u002FSKILL.md",[],[],{"skillNo":70,"name":71,"nameZh":6,"description":72,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":73,"categories":74,"keywords":75,"publishedAt":19},"01M7DVH3HGW47BX3PEF30JV6EJ","mk-vulnerability-scanner","Vulnerability analysis: OWASP Top 10, current threat landscape, supply chain security, attack surface mapping, risk prioritization. Use for security audits. NOT for STRIDE\u002FCSO infra audits (mk:cso).","https:\u002F\u002Fgithub.com\u002Fngocsangyem\u002FMeowKit\u002Fblob\u002Fmain\u002Fkit\u002Fbundles\u002Fcursor\u002Froot\u002F.cursor\u002Fskills\u002Fmk-vulnerability-scanner\u002FSKILL.md",[],[],{"skillNo":77,"name":78,"nameZh":6,"description":79,"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},"01M7DVGZBKMB9WDFH9MFMA99FV","mk-vue","Vue 3 authoring\u002Freview: components, composables, reactivity, Pinia, Pinia Colada, file routing, forms. Composition API + script setup. Auto-activates on .vue files. Deep review: mk:vue-best-practices.","https:\u002F\u002Fgithub.com\u002Fngocsangyem\u002FMeowKit\u002Fblob\u002Fmain\u002Fkit\u002Fbundles\u002Fcursor\u002Froot\u002F.cursor\u002Fskills\u002Fmk-vue\u002FSKILL.md",[],[],{"skillNo":84,"name":85,"nameZh":6,"description":86,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":87,"categories":88,"keywords":89,"publishedAt":19},"01M7DVGSE12AA1M7599HB8HRCT","mk-vue-testing-best-practices","Vue 3 testing best-practices review: Vitest + Vue Test Utils for components, composables, Pinia, Router, async\u002FSuspense, forms, a11y, plus Playwright E2E strategy. Advisory only.","https:\u002F\u002Fgithub.com\u002Fngocsangyem\u002FMeowKit\u002Fblob\u002Fmain\u002Fkit\u002Fbundles\u002Fcursor\u002Froot\u002F.cursor\u002Fskills\u002Fmk-vue-testing-best-practices\u002FSKILL.md",[],[]]