house_rules Skill (ai-integration-01 Compliant) and Auto-Updater
Why
The review findings and ledger data are only useful if they influence future coding sessions. The house_rules.md skill converts accumulated violation patterns into a compact, self-updating context file (≤35 lines) injected into every Claude Code session and every n8n F-2 container. Without this feedback loop, each session starts cold with no awareness of which patterns the codebase consistently violates.
This change follows the ai-integration-01 open Agent Skills standard — skill file at skills/specfact-code-review/SKILL.md in specfact-cli, installable via specfact ide skill install --type code-review.
ALIGNMENT NOTES (from overlap analysis):
- Skill file lives at
skills/specfact-code-review/SKILL.md (not .claude/skills/) — follows ai-integration-01 standard
- Does NOT modify
CLAUDE.md directly — that is owned by ai-integration-03
- house_rules content is the dynamic body; delivery mechanism follows ai-integration-01
What Changes
- NEW:
rules/updater.py — reads ledger findings (last 20 runs), counts violation frequency, surfaces rules with >=3 hits, prunes rules with 0 hits for 10+ consecutive runs, enforces ≤35 line cap, increments version header
- NEW:
specfact code review rules show — prints current house_rules skill content
- NEW:
specfact code review rules update — re-derives house_rules from ledger + recent findings
- NEW:
specfact code review rules init — creates default SKILL.md for new projects
- NEW:
skills/specfact-code-review/SKILL.md (in specfact-cli repo) — ai-integration-01 compliant YAML frontmatter + Markdown body (DO/DON'T rules + TOP VIOLATIONS section)
- NEW:
.cursor/rules/house_rules.mdc mirror (auto-maintained)
- CONSTRAINT: Skill file never exceeds 35 lines (hard cap with assertion)
- CONSTRAINT: No direct
CLAUDE.md modifications (ai-integration-03 owns that path)
- NEW: Unit tests for
updater.py (TDD-first)
Capabilities
New Capabilities
house-rules-skill: ai-integration-01 compliant skill file with auto-updating DO/DON'T rules derived from ledgerrules-commands: specfact code review rules show|update|init subcommandshouse-rules-updater: Algorithm to surface frequent violations and prune resolved rules
Modified Capabilities
specfact-ide-skill-install (ai-integration-01): extended with --type code-review selector (soft dependency)
Impact
- Depends on
code-review-01-module-scaffold, code-review-06-reward-ledger
- Soft dependency on
ai-integration-01 for specfact ide skill install --type code-review; skill file is usable standalone before ai-integration-01 ships
- Documentation: Add rules commands and house_rules skill usage to
docs/modules/code-review.md
OpenSpec Change Proposal: code-review-07-house-rules-skill
house_rules Skill (ai-integration-01 Compliant) and Auto-Updater
Why
The review findings and ledger data are only useful if they influence future coding sessions. The
house_rules.mdskill converts accumulated violation patterns into a compact, self-updating context file (≤35 lines) injected into every Claude Code session and every n8n F-2 container. Without this feedback loop, each session starts cold with no awareness of which patterns the codebase consistently violates.This change follows the
ai-integration-01open Agent Skills standard — skill file atskills/specfact-code-review/SKILL.mdin specfact-cli, installable viaspecfact ide skill install --type code-review.ALIGNMENT NOTES (from overlap analysis):
skills/specfact-code-review/SKILL.md(not.claude/skills/) — follows ai-integration-01 standardCLAUDE.mddirectly — that is owned byai-integration-03What Changes
rules/updater.py— reads ledger findings (last 20 runs), counts violation frequency, surfaces rules with >=3 hits, prunes rules with 0 hits for 10+ consecutive runs, enforces ≤35 line cap, increments version headerspecfact code review rules show— prints current house_rules skill contentspecfact code review rules update— re-derives house_rules from ledger + recent findingsspecfact code review rules init— creates defaultSKILL.mdfor new projectsskills/specfact-code-review/SKILL.md(in specfact-cli repo) — ai-integration-01 compliant YAML frontmatter + Markdown body (DO/DON'T rules + TOP VIOLATIONS section).cursor/rules/house_rules.mdcmirror (auto-maintained)CLAUDE.mdmodifications (ai-integration-03 owns that path)updater.py(TDD-first)Capabilities
New Capabilities
house-rules-skill: ai-integration-01 compliant skill file with auto-updating DO/DON'T rules derived from ledgerrules-commands:specfact code review rules show|update|initsubcommandshouse-rules-updater: Algorithm to surface frequent violations and prune resolved rulesModified Capabilities
specfact-ide-skill-install(ai-integration-01): extended with--type code-reviewselector (soft dependency)Impact
code-review-01-module-scaffold,code-review-06-reward-ledgerai-integration-01forspecfact ide skill install --type code-review; skill file is usable standalone before ai-integration-01 shipsdocs/modules/code-review.mdOpenSpec Change Proposal:
code-review-07-house-rules-skill