Apply this skill whenever creating, refining, splitting, consolidating, renaming, or auditing any agent skill under the host project's skill root.
This skill governs authoring discipline for skills written in the agentskills.io format. For the host project's active skill inventory and topic-to-skill routing, defer to the project's master skill index and the directory listing under the skill root.
Scoping and MECE
See scoping-and-mece.md for:
- choosing a coherent skill boundary, skill name, split, consolidation, or source-of-truth location
- checking overlap with neighboring skills before adding new guidance
- using section length and topic growth as signals for restructuring
Frontmatter and Naming
See frontmatter-and-naming.md for:
- creating or editing discovery-critical
SKILL.mdfrontmatter - choosing the skill directory name and keeping it aligned with the
namefield - porting or preserving host-project harness fields
Description Writing
See description-writing.md for:
- drafting, trimming, or auditing the
descriptionfield - making discovery metadata say both what the skill covers and when to apply it
- adding likely user phrasings and symptom-based triggers without over-broadening the skill
Body Content Style
See body-content-style.md for:
- writing or revising substantive skill-body or reference-file sections
- balancing concise topic explanation, examples, and guideline bullets
- placing normative RFC-2119 requirement bullets in detailed reference content rather than parent routing sections
Progressive Disclosure
See progressive-disclosure.md for:
- deciding when a skill should stay single-file or split into
references/ - using the parent routing-section format:
## Topic,See [file.md](./references/file.md) for:, then descriptive situation bullets - keeping parent routing bullets free of RFC-2119-style requirement keywords so they remain routing cues, not duplicated rules
Cross-Referencing and Index Sync
See cross-referencing.md for:
- adding, renaming, moving, deleting, or linking skills and reference files
- choosing one source of truth instead of copying detailed rules across skills
- verifying relative links and keeping the host project's master skill index synchronized
Audit Checklist
See audit-checklist.md for:
- auditing multiple skills or reporting skill-tree quality
- checking inventory, index sync, section anatomy, RFC-2119 bullets, and relative links
- identifying overlap, stale assumptions, orphan references, and missing source-of-truth links