| name | writing-agents-config |
| description | Codemap repo hybrid deltas for .agents/ โ tiers, rule-skill pairing, repo exemplars. Use when creating or reviewing rules/skills in this repo. Read writing-great-skills first for skill vocabulary and authoring principles. |
| disable-model-invocation | true |
Writing agents config
Read first: writing-great-skills โ invocation, information hierarchy, completion criteria, pruning, failure modes.
This repo adds Cursor rules (Tier 1/2/3) on top of a skills-only agent model โ always-on STOP rules where cross-turn non-negotiables beat description-only triggers.
Tier framework: agents-tier-system ยท Layout: agents-first-convention.
Attachment decision tree
- Every turn, non-negotiable? โ Tier 1 (
alwaysApply: true) โ โค40 lines STOP + pointers
- File-scoped pattern? โ Tier 2 (
globs:) โ thin priming; depth in paired skill
- Intent-only workflow? โ Tier 3 or skill-only โ no always-on tax (bug diagnosis, doc sweeps, structural queries)
- User slash-command only? โ
disable-model-invocation: true wrapper โ model skill(s)
Pairing checklist
| Check | Pass when |
|---|
| Priming size | Rule ~10โ40L; procedures in skill or sibling |
| Bidirectional | Rule โ skill in Reference; skill โ rule where paired |
| No duplication | SSOT in skill; rule has STOP rows or pointers only |
| Inventory drift | Tier-1 set discovered via the frontmatter audit in agents-tier-system rule โ not hardcoded name lists in README/skill |
| Triggers | Per writing-great-skills โ one branch per intent in description |
SKILL.md size tiers
Rules cap at โค40L; skills use progressive disclosure (writing-great-skills sprawl). Line count is a hygiene signal, not a hard gate โ split for structure, not to hit a number.
| Lines | Target |
|---|
| โค60 | Default for new skills |
| 61โ120 | OK for one single-path workflow with checkable completion criteria |
| >120 | Audit โ review for sprawl; disclose to siblings when structure says split |
| >140 | Split unless an explicit bulk reference hub (FULL-GUIDE.md, WORKFLOW.md, REFERENCE.md) |
Split when: branching paths, reference encyclopedia in SKILL.md, premature-completion risk, or sediment โ disclose to siblings (WORKFLOW.md, REFERENCE.md, LANGUAGE.md, PROSE.md).
Don't split when: every line is ordered steps for one branch; splitting forces a multi-file hop mid-checklist.
find .agents/skills -name SKILL.md -exec sh -c 'n=$(wc -l < "$1"); [ "$n" -gt 120 ] && printf "%3d %s\n" "$n" "$1"' _ {} \;
Repo exemplars
consumer-surfaces (Tier 1, always-on), tracer-bullets (rule), docs-governance (Tier 2 globs + skill + docs-lifecycle-sweep), architecture-priming โ improve-codebase-architecture + audit-pr-architecture, codemap (skill-only), harden-pr (SKILL + LEDGER.md), authoring-discipline (rule + PROSE.md), verify-after-each-step (slim rule + skill), agents-tier-system (rule + skill), pr-comment-fact-check (rule + skill).
User-only router
ask-agents โ grill-me, grill-with-docs, writing-great-skills, writing-agents-config. Wrappers โค10 lines.
Codemap-specific tradeoffs
- Tier-2 attach โ
agents-tier-system (agents/** + cursor/**); docs-governance (docs/**, .agents/**); architecture-priming when authoring structural refactors.
- Intent-only skills โ
improve-codebase-architecture, domain-modeling, docs-lifecycle-sweep, diagnosing-bugs, tdd, pr-comment-fact-check, harden-pr, codemap, audit-pr-architecture. No glob (no per-file tax).
- Ship policy โ
templates/agent-content/** is served live (CLI/MCP/HTTP); templates/agents/** is copied by codemap agents init. Consumer surfaces describe behavior only (consumer-surfaces).
- Symlink convention โ source under
.agents/; .cursor/rules/<name>.mdc and .cursor/skills/<name> are symlinks only.
- Cross-skill links โ relative
../skill/SKILL.md and sibling files resolve reliably in Cursor agents. Relative links inside .agents/ are an intentional delta, not drift.
- Indexer-first exploration โ skills that explore structure prefer
codemap query over grep; the codemap rule is Tier 1 always-on.
Anti-patterns
- โ Fat always-on rule with procedures (slim โ skill)
- โ README skill inventories that drift (discover via
ls + frontmatter audit)
- โ Maintainer internals on consumer surfaces (
templates/agent-content, changesets, CLI help)
- โ External-monorepo artifact leakage in examples โ use
src/, index/recipe seams, bun scripts, origin/main
Reference