67%

Docs / spec sync

Goal

Sync all docs to the new graph capability so the generated rule views, graph schema doc, and skill body never drift from the engine’s real capability.

Implementation Sketch

Regenerate the rule views; update intelligence.md graph schema + detector inventory; update the skill to the tier-engine model; /sync-claude for any rule text referencing the old matrix.

Spec References

generator.py output; .claude/rules/intelligence.md; impl-hygiene-review/SKILL.md + phase docs; /sync-claude.

Work Items

  • Regenerate the impl-hygiene-rules generated views (generator run); staleness green.
  • Update intelligence.md: new graph schema (CodeComment nodes, typed dataflow props, new edges) + the new detector inventory + Graph-First detection table.
  • Update /impl-hygiene-review SKILL.md + phase docs to the tier-engine model; remove file-walk references.
  • /sync-claude pass: CLAUDE.md + rules reflect graph-driven detection where any text references the old matrix.

Items

  • Update intelligence.md to the LIVE graph schema: refresh the headline constants to 195K+ symbols / 530K+ CALLS / 11 repos; add the GDS Symbol-prop schema (hotspot_score, pagerank_score, betweenness_score, is_hub, is_articulation, is_dead_code_candidate, coverage_status, scattered_knowledge, phase_order_violation, community_leiden) + edge inventory (MISSING_ABSTRACTION, DataFlowNode def/use base, DEF_USE, IN_COMMUNITY, Comment); add the seam_scan detector inventory + 5-tier classification. Use the LIVE labels (Comment, not CodeComment; DataFlowNode def/use, no operation_kind) and document the comment-for-ori + typed-operation enrichment as capability the plan adds but NOT yet joined for ori in the live snapshot (mirror the engine’s graceful-degradation honesty); never transcribe the sketch’s aspirational CodeComment/operation_kind names verbatim.
  • Reconcile the /impl-hygiene-review phase docs (phase-0..6) to the descriptor-driven tier-engine model and remove residual matrix / file x batch MODEL prose where it describes the engine’s primary path. PRESERVE the graph-down degradation fallback (the legacy file x batch path is the mission-mandated graceful-degradation surface: ‘graph-down NEVER halts; degrade to the existing path’) — frame it as the fallback, never as the model. SKILL.md is already on the tier-engine model (sibling runtime-restructure collapsed the matrix); confirm it and adjust only stray prose that contradicts the engine.

Fresh intel (regenerated)

{ “schema_version”: 1, “target”: { “kind”: “plan-section”, “ref”: “graph-driven-hygiene/s-5ca1550d” }, “generated_at”: “2026-06-22T07:57:12.816254+00:00”, “graph_state”: { “head_sha”: “9e4174b6”, “last_code_import_at”: “2026-06-22T06:40:04.905Z”, “embedding_stale”: false, “insights_stale”: false, “cpg_stale”: false }, “surfaces”: {}, “summary”: “intel-package[plan-section:graph-driven-hygiene/s-5ca1550d] agent-authored dossier”, “degraded”: false, “dossier”: { “objective_symbols”: [], “tiers”: {}, “agent_authored”: true, “difficulty”: { “class”: “routine”, “research_online”: false, “research_mode”: “auto”, “signals”: [], “plan_dir”: “/home/eric/projects/ori_lang/plans/graph-driven-hygiene” } } }

(full dossier: docs-sync—s-5ca1550d.intel.json)