skill-author

name	skill-author
description	Create, repair, review, archive, or merge repo SKILL.md source; handle legacy create-skill requests, trigger text, invocation lanes, routing, owner paths, or dependencies.

Skill Author

Intent Classification

Classify from args top to bottom and proceed. Read CONTEXT.md only when creating a new skill or when vocabulary matters. Show the menu only for no args, missing target on target-specific routes, or ambiguous intent.

Signal	Route	References to open
create-skill named as the skill, path, or handoff target	Normalize target to skill-author, then continue classification from the next matching row; patch live references instead of adding an alias unless active-reference evidence proves a bridge is needed	skills/skill-author/SKILL.md; references/skill-frontmatter-gate.md only when route evidence or bridge decision changes
Target skill + review/audit/check + patch/fix/apply	Review, then patch requested findings	Target SKILL.md; references/skill-design-decision-runbook.md; references/skill-review-rubric.md, then open the smallest branch reference for requested findings
Target skill + review/audit/check + workflow fitness / working workflow / skill-authoring workflow	Review workflow fitness	Target SKILL.md; references/skill-design-decision-runbook.md; references/skill-review-rubric.md; references/skill-workflow-fitness-probes.md; smallest relevant branch references read-only
Target skill + review/audit/check	Review target skill	Target SKILL.md; references/skill-design-decision-runbook.md, then references/skill-review-rubric.md
Target skill + description/trigger/frontmatter/invocation lane/model lane/self invocation	Fix trigger/frontmatter	Target SKILL.md; references/skill-frontmatter-gate.md
Target skill + body/headings/first screen/run card/no-args/next action/examples	Fix body shape	Target SKILL.md; references/skill-body-shape-gate.md; references/skill-io-shape-examples.md when heading shape is unclear
Target skill + owner path/contract/reference/dependency	Fix owner paths	Target SKILL.md; references/skill-owner-path-gate.md; references/skill-dependency-rules.md when dependency behavior changes
Target skill + safety/gotcha/private/destructive/auth/side effect	Fix safety gate	Target SKILL.md; references/skill-safety-gate.md
Target skill + verification/check/test/YAML/handoff	Fix verification	Target SKILL.md; references/skill-verification-gate.md
Target skill + fix/heal/repair/improve/patch	Classify edit branch	Target SKILL.md; references/skill-design-decision-runbook.md, then open the smallest branch reference
"create" / "new skill" + Runtime / CLI / helper command / machine output / durable write / external side effect / repair envelope	Create a runtime-backed skill; design runtime first, then shape SKILL.md through frontmatter and body gates before writing source	CONTEXT.md; references/skill-design-decision-runbook.md; references/skill-safety-gate.md; references/agent-native-skill-design.md; references/runtime-portability.md; skills/cli-author/SKILL.md; references/skill-frontmatter-gate.md; references/skill-body-shape-gate.md; references/skill-io-shape-examples.md only when heading shape is unclear
Runtime / CLI / helper command / machine output / durable write / external side effect / repair envelope	Add runtime behavior to an existing skill or runtime owner	references/skill-design-decision-runbook.md; references/skill-safety-gate.md; references/agent-native-skill-design.md; references/runtime-portability.md; skills/cli-author/SKILL.md
"create" / "new skill"	Create a tiny prose skill by default; add owner, safety, or runtime gates only when evidence earns them	CONTEXT.md; references/skill-design-decision-runbook.md; references/skill-frontmatter-gate.md; references/skill-body-shape-gate.md; references/skill-io-shape-examples.md only when heading shape is unclear
mcporter / MCP via CLI / MCP config / thin MCP skill / server alias / tool schema	Add MC Porter skill guidance	Target SKILL.md when fixing; references/mcporter-skill-design.md; branch gate only when the SKILL.md body changes
Blocked / degraded / dependency	Check dependency behavior	references/skill-dependency-rules.md
Archive / merge / retire	Archive or merge	references/archive-cleanup.md; references/consolidation-map.md
Research / import / handover	Import external input	references/research-portability.md; references/community-skill-research-sources.md
Context / where to save	Route to context advisor	skills/context-advisor/SKILL.md
No args / missing target on target-specific route	Show menu below	-
Ambiguous	Show menu below	-

No-args or ambiguous menu

Present only for no args, missing target on target-specific routes, or when intent classification cannot pick a route:

1. Fix, heal, or repair a skill - name the target, open its SKILL.md.
2. Review an existing skill - read target + runbook review-only branch + review rubric, return findings.
3. Create a new skill - read CONTEXT.md; default to the tiny prose path unless runtime, owner, or safety evidence earns more.
4. Unsure - read CONTEXT.md, then pick the closest route above or stop.

Run Card

• Scope: create, repair, review, archive, or merge repo skill source files.
• First safe action: show the menu for no args, missing target on target-specific routes, or ambiguous intent; otherwise classify intent from args and open only the references for that route.
• Mode defaults: review-only returns findings; mixed review plus patch edits only requested findings; create, fix, heal, repair, or patch edits source.
• Review boundary: do not patch during review unless the user asks for edits.
• Thin-router gate: keep SKILL.md a thin router for the current step only.
• Pruning gate: apply the deletion test; headings are options, not a checklist.
• Fallback: stop when owner path, input/output shape, write authority, or target skill is unclear.

Owner Anchors

• Branch index: references/skill-design-decision-runbook.md.
• Review rubric: references/skill-review-rubric.md.