writing-agents-config

name	writing-agents-config
description	Layers repo hybrid deltas for .agents/ — 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.

Writing agents config (layers hybrid deltas)

Read first: writing-great-skills — 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 — always-on STOP rules where cross-turn non-negotiables beat description-only triggers.

Tier framework: agents-tier-system · Layout: agents-first-convention.

Attachment decision tree (layers)

1. Every turn, non-negotiable? → Tier 1 (alwaysApply: true) — ≤40 lines STOP + pointers
2. File-scoped pattern? → Tier 2 (globs:) — thin priming; depth in paired skill
3. Intent-only workflow? → Tier 3 or skill-only — no always-on tax (bug diagnosis, doc sweeps)
4. User slash-command only? → disable-model-invocation: true wrapper → model skill(s)

Pairing checklist (layers)

Check	Pass when
Priming size	Rule ~10–40L; procedures in skill or sibling
Bidirectional	Rule → skill in Reference; skill → 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 — not hardcoded name lists in README/skill
Triggers	Per writing-great-skills — one branch per intent in description

SKILL.md size tiers

Rules cap at ≤40L; skills use progressive disclosure (writing-great-skills sprawl). Line count is a hygiene signal, not a hard gate — split for structure, not to hit a number.

Lines	Target
≤60	Default for new skills
61–120	OK for one single-path workflow with checkable completion criteria
>120	Audit — 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 — 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.

# Review tail (>120L) — split only when table above says so
find .agents/skills -name SKILL.md -exec sh -c 'n=$(wc -l < "$1"); [ "$n" -gt 120 ] && printf "%3d %s\n" "$n" "$1"' _ {} \;

Repo exemplars

tracer-bullets (rule, single file), harden-pr (SKILL + WORKFLOW.md + LEDGER.md), docs-governance (SKILL + LIFECYCLE.md), improve-codebase-architecture (SKILL + LANGUAGE.md + REFERENCE.md), authoring-discipline (rule + PROSE.md), verify-after-each-step (slim rule + skill), agents-tier-system (rule + skill).

User-only router

ask-agents — grill-me, grill-with-docs, teach, writing-great-skills, writing-agents-config, minimum-diff. Wrappers ≤10 lines.

Layers-specific tradeoffs

• Tier-2 attach — agents-tier-system (.agents/rules/**, .agents/skills/**, .cursor/rules/**, .cursor/skills/**); docs-governance-priming (docs/**, .agents/**, .cursor/**).
• Intent-only skills — improve-codebase-architecture, domain-modeling, docs-lifecycle-sweep, diagnosing-bugs, tdd, pr-comment-fact-check, harden-pr. No glob (no per-file tax).
• Cross-skill links — relative ../skill/SKILL.md and sibling files resolve reliably in Cursor agents. Relative links inside .agents/ are an intentional delta, not drift.
• Small-lib surface — no per-feature docs/ subtrees, no UI-component skills, no API-client/codegen skills. Don't add a skill governing tooling this repo lacks (a rule referencing a nonexistent script is worse than no rule).

Anti-patterns (layers)

• ❌ Fat always-on rule with procedures (slim → skill)
• ❌ README skill inventories that drift (discover via ls + frontmatter audit)
• ❌ Brittle anchors in committed source (authoring-discipline)
• ❌ External-monorepo artifact leakage in examples — use packages/core/src/, packages/<fw>/src/, the package-level core/adapter seam, bun scripts, origin/main

Reference

• Skill authoring SSOT: writing-great-skills
• Tiers + audit: agents-tier-system
• Docs lifecycle: docs-governance