[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7F2CFVDX317GZF4D1BYC5W5":3,"skill-same:01M7F2CFVDX317GZF4D1BYC5W5":68},{"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":36,"commitSha":37,"skillDirectory":38,"forksCount":39,"repositoryHtmlUrl":40,"repositoryHomepageUrl":6,"ownerHtmlUrl":41,"ownerBlogUrl":6,"ownerDescription":42,"updatedAt":43,"install":44,"resources":57},"01M7F2CFVDX317GZF4D1BYC5W5","omh-triage",null,"Multi-role consensus triage of an issue backlog.","lza6\u002FArmor-piercing-system-prompts","lza6","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F128880206?v=4","01M7EB3K7WPX5T28EJ2NVHE6EG","Chinese-tingfeng","01M7EB3K97E1KQ16QWKJMFNGX8",11,"MIT","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F05-技能与代理开发\u002Fomh-triage\u002FSKILL.md",[],[],"2026-11-16 10:05:34",false,"---\nname: omh-triage\ndescription: \"Multi-role consensus triage of an issue backlog.\"\nversion: 0.1.0\nmetadata:\n  hermes:\n    tags: [triage, multi-agent, consensus, backlog]\n    category: omh\n    requires_toolsets: [terminal, omh, delegation]\n---\n\n# OMH Triage — Consensus Backlog Grooming\n\n> **v0.1 status:** This skill is **deliberately small**. It ships with two roles (Maintainer, Skeptic) and is being battle-tested before more roles are added. The full design hypothesis (Maintainer \u002F Operator \u002F Architect \u002F Member-advocate \u002F Skeptic) lives in [witt3rd\u002Foh-my-hermes#9](https:\u002F\u002Fgithub.com\u002Fwitt3rd\u002Foh-my-hermes\u002Fissues\u002F9). The next concrete milestone (Operator role for version-cut planning pressure) is tracked at [witt3rd\u002Foh-my-hermes#11](https:\u002F\u002Fgithub.com\u002Fwitt3rd\u002Foh-my-hermes\u002Fissues\u002F11). Roles will be added as lived rounds surface real tensions that the current set cannot represent. The discipline here is *learn through use, then expand* — the inverse of authoring all five roles up-front against imagined friction.\n\n## When to Use\n\n- Periodically grooming an issue backlog (default cadence: before each version cut, when ≥10 issues have accumulated since the last pass, or when a major refactor\u002Fmigration just landed and may have moved issue premises)\n- After a fast-moving development arc (multiple migrations or refactors in days\u002Fweeks) that may have superseded several open issues\n- When the user says: \"triage\", \"groom the backlog\", \"walk the issues\", \"what's stale\"\n\n## When NOT to Use\n\n- A single ambiguous issue that needs design (use `omh-deep-interview` then `omh-ralplan`)\n- A backlog that's never been groomed and has hundreds of issues (do a manual first-pass cull first; this skill is for keeping a curated backlog honest, not for archaeology on a rotted one)\n- Closing during active dev (don't mix the two-mode discipline — capture during dev, decide during grooming; see the host project's `docs\u002Ftriage\u002FPROCESS.md` if it has one)\n\n## Prerequisites\n\n- A target repo with an issue backlog (GitHub Issues; pluggable for other backlogs eventually)\n- Optional: a GitHub Project for version + cluster + status targeting (the run will write back to it)\n- Optional: prior triage docs under `docs\u002Ftriage\u002F` for delta-since-last-pass framing\n- The `delegate_task` tool must be available\n\n## Procedure\n\n### Phase 0: Ground-truth pre-flight\n\nBefore dispatching any roles, the orchestrator gathers:\n\n1. **Repo HEAD.** Note the commit ref. Verdicts will anchor against this.\n2. **Issue inventory.** `gh issue list --state open --limit \u003CN>`. Note count, label distribution, age distribution.\n3. **Recent migrations \u002F refactors.** Read commit log for the last N weeks. The orchestrator must know what surfaces have moved before dispatching — otherwise the Maintainer will discover this in-flight and waste a round.\n4. **Existing project board state** (if any). Read target version, cluster, status fields for context.\n5. **Prior triage doc** (if any). The most recent `docs\u002Ftriage\u002FYYYY-MM-DD-*.md` is delta-since-last-pass anchor.\n\nPhase 0's output is a ~500-word context package that all subagents will receive. Same shape as `omh-ralplan` Phase 0.\n\n### Phase 1: Role passes (sequential, parallel where possible)\n\n**v0.1 ships two roles.** Future versions will add Operator, Architect, Member-advocate.\n\n#### Step 1 — Maintainer pass\n\n```\ndelegate_task(\n    goal=\"[omh-role:triage-maintainer] Vet each issue in the attached inventory against current code. Produce a per-issue verdict (stale\u002Frecast\u002Flive\u002Fpartial-stale\u002Fout-of-scope) with code-anchored reasoning. Repo HEAD: \u003Cref>. Inventory: \u003Clist>\",\n    context=\"# Project context\\n\\n\u003Ccontext-package>\\n\\n# Inventory\\n\\n\u003Cfull issue bodies, not just titles>\"\n)\n```\n\nThe Maintainer's pass is **per-issue**. Output is structured (one block per issue with verdict + anchor + reasoning + close-comment-or-recast-spec).\n\n#### Step 2 — Skeptic pass (after Maintainer; needs Maintainer's verdicts as input)\n\n```\ndelegate_task(\n    goal=\"[omh-role:triage-skeptic] For each LIVE issue from the Maintainer pass, apply skeptical pressure: keep\u002Fdrop\u002Fdedup\u002Frefile-smaller\u002Fwait-for-recurrence. Stale\u002Fout-of-scope issues do not need skeptical review.\",\n    context=\"# Project context\\n\\n\u003Ccontext-package>\\n\\n# Maintainer verdicts\\n\\n\u003CMaintainer's structured output>\\n\\n# Full inventory (for cross-reference)\\n\\n\u003Cfull issue bodies>\"\n)\n```\n\nThe Skeptic only reviews issues the Maintainer marked live or recast. Stale\u002Fout-of-scope are already moving to close; double-pressure on them wastes the role's attention.\n\n#### Step 3 — Orchestrator distillation\n\nThe orchestrator combines Maintainer + Skeptic verdicts. **The matrix below is authoritative** — every Maintainer × Skeptic combination resolves to a specific disposition. Combinations not enumerated here are the only cases that escalate to user as a true conflict.\n\nThe Maintainer verdict `partial-stale` (issue contains multiple sub-claims; some stale, some live) is treated identically to `recast` for matrix purposes — both signal \"body needs surgery, possibly extensive.\" Maintainer output should still distinguish them so the recast spec can name which sub-claims are stale vs live, but the disposition resolution is the same.\n\n| Maintainer | Skeptic | Resolved disposition |\n|------------|---------|----------------------|\n| stale | (not run) | close — Maintainer pointer-comment |\n| out-of-scope | (not run) | close — refile-target named |\n| recast \u002F partial-stale | keep | recast body, keep open |\n| recast \u002F partial-stale | drop \u002F wait-for-recurrence | close — Skeptic pointer-comment (the body needs work AND the underlying friction may not be load-bearing; combined verdict is \"not worth the surgery\") |\n| recast \u002F partial-stale | dedup | close + comment-on-covering-issue |\n| recast \u002F partial-stale | refile-smaller | close + reopen-as-smaller |\n| live | keep | live |\n| live | drop \u002F wait-for-recurrence | close — Skeptic pointer-comment |\n| live | dedup | close + comment-on-covering-issue |\n| live | refile-smaller | close + reopen-as-smaller |\n\nThe matrix is exhaustive over the verdict types in this skill (v0.1). When future versions add roles or new verdict types, the matrix expands; conflicts surface only when a *new* combination doesn't have a resolution.\n\n**The escalation case** (rare in v0.1): if a Skeptic returns `needs lived signal` (the role's escape hatch when verdict cannot be determined) — or any other unresolved value the role catalog doesn't enumerate — that's a true escalation. The orchestrator surfaces these to user as specific decision questions. The orchestrator's job is to distill, not to decide for the user when the matrix doesn't resolve.\n\n### Phase 2: Output\n\nThe orchestrator produces:\n\n1. **Triage doc** at `docs\u002Ftriage\u002FYYYY-MM-DD-\u003Cdescriptor>.md` (in the host repo, not in this skill repo) using the host project's template. If no template exists, use the structure documented in `references\u002Ftriage-doc-template.md`.\n2. **Verdict execution plan** — list of closures (with body files staged at `\u002Ftmp\u002FcloseN.md`), recasts (with proposed body diffs), refile-smaller drafts.\n3. **Project board updates** (if a project is configured) — target-version moves, cluster updates, status changes.\n\nThe orchestrator **stages, does not execute** by default. The user sign-off gate happens between Phase 2 and execution. See `references\u002Forchestrator-review-template.md`.\n\n### Phase 3: Execution (user-gated)\n\nAfter user approval:\n\n1. Post pointer-comments and close stale\u002Fdropped issues\n2. Post recast PRs (or comments with proposed body changes)\n3. Update project board fields\n4. Commit the triage doc with appropriate authorship\n\n## Roles in v0.1\n\n| Role | Reference | Pressure |\n|------|-----------|----------|\n| Triage Maintainer | `references\u002Frole-triage-maintainer.md` | Code-anchored ground truth: is the premise live? |\n| Triage Skeptic | `references\u002Frole-triage-skeptic.md` | Pruning: does this earn its slot? |\n\nRoles **planned for v0.2+** (witt3rd\u002Foh-my-hermes#9):\n\n- **Operator** — version-cut planning: smallest defensible vN+1\n- **Architect** — cross-cutting: which issues are symptoms of one root\n- **Member-advocate** — lived experience: what hurts the user\u002Fprincipal\u002Fbeing now\n\nThese will be added when lived rounds surface tensions the current two cannot represent. Resist authoring them ahead of evidence.\n\n## Output Quality Bar\n\nA run is done when:\n\n1. Every open issue has a Maintainer verdict (no skipped issues)\n2. Every live\u002Frecast issue has a Skeptic verdict\n3. Every stale\u002Fdropped issue has a pointer-comment ready to post\n4. Every recast issue has a specific body-surgery spec (not \"rewrite the body\")\n5. Conflicts are escalated to user, not silently resolved\n6. The triage doc names what this round did NOT do (deferred work)\n\n## Pitfalls (v0.1, will grow)\n\n### T1 — Don't dispatch Maintainer without ground-truth Phase 0\n\nMaintainer's job is to anchor against current code. If the orchestrator hasn't already mapped recent migrations\u002Frefactors, the Maintainer discovers them in-flight and wastes attention re-deriving what Phase 0 should have given.\n\n### T2 — Skeptic only reviews live + recast\n\nIssues the Maintainer marked stale or out-of-scope are already moving to close. Sending them to Skeptic produces noise (Skeptic will say \"drop\" on stale issues, which adds nothing).\n\n### T3 — Stage, don't execute by default\n\nTriage closes are durable on the issue tracker. Always stage closures with body files at `\u002Ftmp\u002FcloseN.md` and let the user approve before executing the actual close commands. The skill that runs the close after approval should loop in shell with `--comment \"$(cat \u002Ftmp\u002FcloseN.md)\"` (multi-paragraph quoted bodies through `terminal()` from `execute_code` fail silently — lesson from janus round 1).\n\n### T4 — The matrix resolves; only un-enumerated cases escalate\n\nThe verdict-combination matrix in Phase 1 Step 3 is **authoritative** — `live` + `drop` resolves to \"close — Skeptic pointer-comment\", not to \"user conflict.\" The matrix encodes our considered judgment that \"Maintainer says live, Skeptic says drop\" means *the issue is real but not worth the slot it consumes*; closing with a clear pointer-comment that invites refile-on-recurrence is the correct disposition, not a user-decision question.\n\nTrue escalations are: the Skeptic returns `needs lived signal` (escape hatch), the Maintainer returns `needs investigation` (escape hatch), or some future role's verdict produces a combination the matrix doesn't enumerate. These are rare in v0.1 and surface to user as specific decision questions, not as \"there are some conflicts.\"\n\nThe dispatcher should also feel free to **override the matrix** for individual issues when their own bar applied to the verdict produces a different read — but that's a dispatcher-judgment call, not a matrix conflict. Note overrides explicitly in the orchestrator review (`omh-ralplan-driver` P25 — skepticism over deference applies here too).\n\n### T5 — A round produces a durable doc\n\nThe triage doc under `docs\u002Ftriage\u002FYYYY-MM-DD-*.md` is the round's audit trail. It names what was decided, what was deferred, and what the round taught. A round without a doc didn't happen.\n\n### T6 — Resist the pressure to ship missing roles up-front\n\nThis skill is v0.1 deliberately. The OMH#9 design names five roles; only two ship today. Adding Operator \u002F Architect \u002F Member-advocate before lived rounds surface real tensions for them is the \"imagined friction\" failure mode. Wait for evidence.\n\n## Pairs With\n\n- `omh-triage-driver` — the dispatcher's playbook (when to invoke, how to drive, what good output looks like)\n- `omh-ralplan-driver` — analogous skill for the design layer; the discipline patterns are similar\n- `omh-deep-interview` — for ambiguous issues that need clarification before triage can decide\n\n## See Also\n\n- [`witt3rd\u002Foh-my-hermes#9`](https:\u002F\u002Fgithub.com\u002Fwitt3rd\u002Foh-my-hermes\u002Fissues\u002F9) — design tracking issue\n- `references\u002Frole-triage-maintainer.md`\n- `references\u002Frole-triage-skeptic.md`\n","# OMH Triage — Consensus Backlog Grooming\n\n> **v0.1 status:** This skill is **deliberately small**. It ships with two roles (Maintainer, Skeptic) and is being battle-tested before more roles are added. The full design hypothesis (Maintainer \u002F Operator \u002F Architect \u002F Member-advocate \u002F Skeptic) lives in [witt3rd\u002Foh-my-hermes#9](https:\u002F\u002Fgithub.com\u002Fwitt3rd\u002Foh-my-hermes\u002Fissues\u002F9). The next concrete milestone (Operator role for version-cut planning pressure) is tracked at [witt3rd\u002Foh-my-hermes#11](https:\u002F\u002Fgithub.com\u002Fwitt3rd\u002Foh-my-hermes\u002Fissues\u002F11). Roles will be added as lived rounds surface real tensions that the current set cannot represent. The discipline here is *learn through use, then expand* — the inverse of authoring all five roles up-front against imagined friction.\n\n## When to Use\n\n- Periodically grooming an issue backlog (default cadence: before each version cut, when ≥10 issues have accumulated since the last pass, or when a major refactor\u002Fmigration just landed and may have moved issue premises)\n- After a fast-moving development arc (multiple migrations or refactors in days\u002Fweeks) that may have superseded several open issues\n- When the user says: \"triage\", \"groom the backlog\", \"walk the issues\", \"what's stale\"\n\n## When NOT to Use\n\n- A single ambiguous issue that needs design (use `omh-deep-interview` then `omh-ralplan`)\n- A backlog that's never been groomed and has hundreds of issues (do a manual first-pass cull first; this skill is for keeping a curated backlog honest, not for archaeology on a rotted one)\n- Closing during active dev (don't mix the two-mode discipline — capture during dev, decide during grooming; see the host project's `docs\u002Ftriage\u002FPROCESS.md` if it has one)\n\n## Prerequisites\n\n- A target repo with an issue backlog (GitHub Issues; pluggable for other backlogs eventually)\n- Optional: a GitHub Project for version + cluster + status targeting (the run will write back to it)\n- Optional: prior triage docs under `docs\u002Ftriage\u002F` for delta-since-last-pass framing\n- The `delegate_task` tool must be available\n\n## Procedure\n\n### Phase 0: Ground-truth pre-flight\n\nBefore dispatching any roles, the orchestrator gathers:\n\n1. **Repo HEAD.** Note the commit ref. Verdicts will anchor against this.\n2. **Issue inventory.** `gh issue list --state open --limit \u003CN>`. Note count, label distribution, age distribution.\n3. **Recent migrations \u002F refactors.** Read commit log for the last N weeks. The orchestrator must know what surfaces have moved before dispatching — otherwise the Maintainer will discover this in-flight and waste a round.\n4. **Existing project board state** (if any). Read target version, cluster, status fields for context.\n5. **Prior triage doc** (if any). The most recent `docs\u002Ftriage\u002FYYYY-MM-DD-*.md` is delta-since-last-pass anchor.\n\nPhase 0's output is a ~500-word context package that all subagents will receive. Same shape as `omh-ralplan` Phase 0.\n\n### Phase 1: Role passes (sequential, parallel where possible)\n\n**v0.1 ships two roles.** Future versions will add Operator, Architect, Member-advocate.\n\n#### Step 1 — Maintainer pass\n\n```\ndelegate_task(\n    goal=\"[omh-role:triage-maintainer] Vet each issue in the attached inventory against current code. Produce a per-issue verdict (stale\u002Frecast\u002Flive\u002Fpartial-stale\u002Fout-of-scope) with code-anchored reasoning. Repo HEAD: \u003Cref>. Inventory: \u003Clist>\",\n    context=\"# Project context\\n\\n\u003Ccontext-package>\\n\\n# Inventory\\n\\n\u003Cfull issue bodies, not just titles>\"\n)\n```\n\nThe Maintainer's pass is **per-issue**. Output is structured (one block per issue with verdict + anchor + reasoning + close-comment-or-recast-spec).\n\n#### Step 2 — Skeptic pass (after Maintainer; needs Maintainer's verdicts as input)\n\n```\ndelegate_task(\n    goal=\"[omh-role:triage-skeptic] For each LIVE issue from the Maintainer pass, apply skeptical pressure: keep\u002Fdrop\u002Fdedup\u002Frefile-smaller\u002Fwait-for-recurrence. Stale\u002Fout-of-scope issues do not need skeptical review.\",\n    context=\"# Project context\\n\\n\u003Ccontext-package>\\n\\n# Maintainer verdicts\\n\\n\u003CMaintainer's structured output>\\n\\n# Full inventory (for cross-reference)\\n\\n\u003Cfull issue bodies>\"\n)\n```\n\nThe Skeptic only reviews issues the Maintainer marked live or recast. Stale\u002Fout-of-scope are already moving to close; double-pressure on them wastes the role's attention.\n\n#### Step 3 — Orchestrator distillation\n\nThe orchestrator combines Maintainer + Skeptic verdicts. **The matrix below is authoritative** — every Maintainer × Skeptic combination resolves to a specific disposition. Combinations not enumerated here are the only cases that escalate to user as a true conflict.\n\nThe Maintainer verdict `partial-stale` (issue contains multiple sub-claims; some stale, some live) is treated identically to `recast` for matrix purposes — both signal \"body needs surgery, possibly extensive.\" Maintainer output should still distinguish them so the recast spec can name which sub-claims are stale vs live, but the disposition resolution is the same.\n\n| Maintainer | Skeptic | Resolved disposition |\n|------------|---------|----------------------|\n| stale | (not run) | close — Maintainer pointer-comment |\n| out-of-scope | (not run) | close — refile-target named |\n| recast \u002F partial-stale | keep | recast body, keep open |\n| recast \u002F partial-stale | drop \u002F wait-for-recurrence | close — Skeptic pointer-comment (the body needs work AND the underlying friction may not be load-bearing; combined verdict is \"not worth the surgery\") |\n| recast \u002F partial-stale | dedup | close + comment-on-covering-issue |\n| recast \u002F partial-stale | refile-smaller | close + reopen-as-smaller |\n| live | keep | live |\n| live | drop \u002F wait-for-recurrence | close — Skeptic pointer-comment |\n| live | dedup | close + comment-on-covering-issue |\n| live | refile-smaller | close + reopen-as-smaller |\n\nThe matrix is exhaustive over the verdict types in this skill (v0.1). When future versions add roles or new verdict types, the matrix expands; conflicts surface only when a *new* combination doesn't have a resolution.\n\n**The escalation case** (rare in v0.1): if a Skeptic returns `needs lived signal` (the role's escape hatch when verdict cannot be determined) — or any other unresolved value the role catalog doesn't enumerate — that's a true escalation. The orchestrator surfaces these to user as specific decision questions. The orchestrator's job is to distill, not to decide for the user when the matrix doesn't resolve.\n\n### Phase 2: Output\n\nThe orchestrator produces:\n\n1. **Triage doc** at `docs\u002Ftriage\u002FYYYY-MM-DD-\u003Cdescriptor>.md` (in the host repo, not in this skill repo) using the host project's template. If no template exists, use the structure documented in `references\u002Ftriage-doc-template.md`.\n2. **Verdict execution plan** — list of closures (with body files staged at `\u002Ftmp\u002FcloseN.md`), recasts (with proposed body diffs), refile-smaller drafts.\n3. **Project board updates** (if a project is configured) — target-version moves, cluster updates, status changes.\n\nThe orchestrator **stages, does not execute** by default. The user sign-off gate happens between Phase 2 and execution. See `references\u002Forchestrator-review-template.md`.\n\n### Phase 3: Execution (user-gated)\n\nAfter user approval:\n\n1. Post pointer-comments and close stale\u002Fdropped issues\n2. Post recast PRs (or comments with proposed body changes)\n3. Update project board fields\n4. Commit the triage doc with appropriate authorship\n\n## Roles in v0.1\n\n| Role | Reference | Pressure |\n|------|-----------|----------|\n| Triage Maintainer | `references\u002Frole-triage-maintainer.md` | Code-anchored ground truth: is the premise live? |\n| Triage Skeptic | `references\u002Frole-triage-skeptic.md` | Pruning: does this earn its slot? |\n\nRoles **planned for v0.2+** (witt3rd\u002Foh-my-hermes#9):\n\n- **Operator** — version-cut planning: smallest defensible vN+1\n- **Architect** — cross-cutting: which issues are symptoms of one root\n- **Member-advocate** — lived experience: what hurts the user\u002Fprincipal\u002Fbeing now\n\nThese will be added when lived rounds surface tensions the current two cannot represent. Resist authoring them ahead of evidence.\n\n## Output Quality Bar\n\nA run is done when:\n\n1. Every open issue has a Maintainer verdict (no skipped issues)\n2. Every live\u002Frecast issue has a Skeptic verdict\n3. Every stale\u002Fdropped issue has a pointer-comment ready to post\n4. Every recast issue has a specific body-surgery spec (not \"rewrite the body\")\n5. Conflicts are escalated to user, not silently resolved\n6. The triage doc names what this round did NOT do (deferred work)\n\n## Pitfalls (v0.1, will grow)\n\n### T1 — Don't dispatch Maintainer without ground-truth Phase 0\n\nMaintainer's job is to anchor against current code. If the orchestrator hasn't already mapped recent migrations\u002Frefactors, the Maintainer discovers them in-flight and wastes attention re-deriving what Phase 0 should have given.\n\n### T2 — Skeptic only reviews live + recast\n\nIssues the Maintainer marked stale or out-of-scope are already moving to close. Sending them to Skeptic produces noise (Skeptic will say \"drop\" on stale issues, which adds nothing).\n\n### T3 — Stage, don't execute by default\n\nTriage closes are durable on the issue tracker. Always stage closures with body files at `\u002Ftmp\u002FcloseN.md` and let the user approve before executing the actual close commands. The skill that runs the close after approval should loop in shell with `--comment \"$(cat \u002Ftmp\u002FcloseN.md)\"` (multi-paragraph quoted bodies through `terminal()` from `execute_code` fail silently — lesson from janus round 1).\n\n### T4 — The matrix resolves; only un-enumerated cases escalate\n\nThe verdict-combination matrix in Phase 1 Step 3 is **authoritative** — `live` + `drop` resolves to \"close — Skeptic pointer-comment\", not to \"user conflict.\" The matrix encodes our considered judgment that \"Maintainer says live, Skeptic says drop\" means *the issue is real but not worth the slot it consumes*; closing with a clear pointer-comment that invites refile-on-recurrence is the correct disposition, not a user-decision question.\n\nTrue escalations are: the Skeptic returns `needs lived signal` (escape hatch), the Maintainer returns `needs investigation` (escape hatch), or some future role's verdict produces a combination the matrix doesn't enumerate. These are rare in v0.1 and surface to user as specific decision questions, not as \"there are some conflicts.\"\n\nThe dispatcher should also feel free to **override the matrix** for individual issues when their own bar applied to the verdict produces a different read — but that's a dispatcher-judgment call, not a matrix conflict. Note overrides explicitly in the orchestrator review (`omh-ralplan-driver` P25 — skepticism over deference applies here too).\n\n### T5 — A round produces a durable doc\n\nThe triage doc under `docs\u002Ftriage\u002FYYYY-MM-DD-*.md` is the round's audit trail. It names what was decided, what was deferred, and what the round taught. A round without a doc didn't happen.\n\n### T6 — Resist the pressure to ship missing roles up-front\n\nThis skill is v0.1 deliberately. The OMH#9 design names five roles; only two ship today. Adding Operator \u002F Architect \u002F Member-advocate before lived rounds surface real tensions for them is the \"imagined friction\" failure mode. Wait for evidence.\n\n## Pairs With\n\n- `omh-triage-driver` — the dispatcher's playbook (when to invoke, how to drive, what good output looks like)\n- `omh-ralplan-driver` — analogous skill for the design layer; the discipline patterns are similar\n- `omh-deep-interview` — for ambiguous issues that need clarification before triage can decide\n\n## See Also\n\n- [`witt3rd\u002Foh-my-hermes#9`](https:\u002F\u002Fgithub.com\u002Fwitt3rd\u002Foh-my-hermes\u002Fissues\u002F9) — design tracking issue\n- `references\u002Frole-triage-maintainer.md`\n- `references\u002Frole-triage-skeptic.md`\n",{"name":5,"version":24,"metadata":25,"description":7},"0.1.0",{"hermes":26},{"tags":27,"category":32,"requires_toolsets":33},[28,29,30,31],"triage","multi-agent","consensus","backlog","omh",[34,32,35],"terminal","delegation","main","d69096e239f2a89aabc2cd078c47664a9d3b1e11","05-技能与代理开发\u002Fomh-triage",3,"https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts","https:\u002F\u002Fgithub.com\u002Flza6","个人vx：Tf00798\r\n个人✈：@TFGY999","2026-10-08 14:29:22",{"source":8,"skillName":5,"commands":45,"prompt":55,"zipAvailable":56},[46,49,52],{"tool":47,"command":48},"npx","npx skills add lza6\u002FArmor-piercing-system-prompts --skill omh-triage",{"tool":50,"command":51},"bunx","bunx skills add lza6\u002FArmor-piercing-system-prompts --skill omh-triage",{"tool":53,"command":54},"pnpm","pnpm dlx skills add lza6\u002FArmor-piercing-system-prompts --skill omh-triage","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F05-技能与代理开发\u002Fomh-triage\u002FSKILL.md",true,[58,64],{"path":59,"resourceType":60,"resourceTypeDesc":61,"sizeBytes":62,"sourceHtmlUrl":63},"references\u002Forchestrator-review-template.md","REFERENCE","参考文档",2423,"https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F05-技能与代理开发\u002Fomh-triage\u002Freferences\u002Forchestrator-review-template.md",{"path":65,"resourceType":60,"resourceTypeDesc":61,"sizeBytes":66,"sourceHtmlUrl":67},"references\u002Ftriage-doc-template.md",1813,"https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F05-技能与代理开发\u002Fomh-triage\u002Freferences\u002Ftriage-doc-template.md",[69,76,81,88,95,102],{"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},"01M7FFG485KY83TD6HSC4YPCMC","obsidian-clipper-template-creator","Guide for creating templates for the Obsidian Web Clipper. Use when you want to create a new clipping template, understand available variables, or format clipped content.","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F07-后端与架构\u002Fobsidian-clipper-template-creator\u002FSKILL.md",[],[],{"skillNo":77,"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":78,"categories":79,"keywords":80,"publishedAt":19},"01M7FFG0T6AZ0STEJPSB3CX84K","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F07-后端与架构\u002Fobsidian-clipper-template-creator-antigravity-awesome-skills-main\u002FSKILL.md",[],[],{"skillNo":82,"name":83,"nameZh":6,"description":84,"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},"01M7FFFY0QEM14SKWPDYGMXM2P","python-api","Programmatic Object-Detection-Metrics API guidance for BoundingBox, BoundingBoxes, Evaluator, enums, coordinate conversion, AP\u002FmAP semantics, and source-style import diagnostics.","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F07-后端与架构\u002Fobject-detection-metrics\u002Fsub-skills\u002Fpython-api\u002FSKILL.md",[],[],{"skillNo":89,"name":90,"nameZh":6,"description":91,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":92,"categories":93,"keywords":94,"publishedAt":19},"01M7FFFVZQZ2GJ0C0TGP7JHD20","file-evaluation","Folder-based PASCAL VOC AP\u002FmAP evaluation from ground-truth and detection text files.","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F07-后端与架构\u002Fobject-detection-metrics\u002Fsub-skills\u002Ffile-evaluation\u002FSKILL.md",[],[],{"skillNo":96,"name":97,"nameZh":6,"description":98,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":99,"categories":100,"keywords":101,"publishedAt":19},"01M7FFFSFNYSE77SJ6PE1CK2DA","object-detection-metrics","Guides legacy Object-Detection-Metrics PASCAL VOC AP\u002FmAP evaluation from detection text folders and direct Python API classes.","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F07-后端与架构\u002Fobject-detection-metrics\u002FSKILL.md",[],[],{"skillNo":103,"name":104,"nameZh":6,"description":105,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":106,"categories":107,"keywords":108,"publishedAt":19},"01M7FFFP0GCK1RVXBCPQ8APBMT","nx-workspace-patterns","Configure and optimize Nx monorepo workspaces. Use when setting up Nx, configuring project boundaries, optimizing build caching, or implementing affected commands.","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F07-后端与架构\u002Fnx-workspace-patterns\u002FSKILL.md",[],[]]