kopi-cli

Six-dimension structured code review framework. Produces severity-ranked findings (CRITICAL/HIGH/MEDIUM/LOW/INFO) with evidence mapped to AC IDs, and a verdict (APPROVE/REQUEST_CHANGES/NEEDS_DISCUSSION). Used by kopi-reviewer agent.

Playwright-based evidence capture protocol. Produces screenshot + HTML + console proof mapped per AC ID. Used by kopi-validate agent to generate pass/fail evidence for the evidence-manifest.

Converts a Figma frame into a structured intent-package with AC confidence scoring, state matrix enumeration, and JP/VI/EN annotation handling. Used by kopi-figspec agent.

Knowledge harvesting — surface, distil, store, and recall tacit business knowledge. Run the Recall → Codebase-first → Grill → Distil → Confirm loop when a task needs a business fact that is not written down (why "予約" is always "reservation", when a feedback is a CR vs a Bug, which flow touches payments). Invokable by kopi-brse, kopi-implement, and any kopi agent that hits a business-knowledge gap. Writes verified facts to knowledge/.

Assemble accumulated project knowledge + context into a compact, reusable context-pack — for the orchestrator to hand to subagents on fan-out (so each subagent does NOT re-gather), or for any agent/member to re-establish context fast on resume. Complements `harvest` (harvest distils NEW tacit facts into knowledge/; context-pack packages EXISTING context — code, knowledge, glossary, structure). Reads knowledge/INDEX.md + glossary + structural signals; never invents new knowledge.

Cross-repository contract analysis. Identifies downstream impact of changes to shared types, API contracts, or shared utilities. Used by kopi-reviewer in cross-repo pattern and when breaking changes are detected.

Root cause analysis for bugs, incidents, and regressions. Produces a structured RCA with reproduction steps, impact assessment, and recommended minimum fix. Used in bug-triage and hotfix patterns.

Manual, on-demand review (NOT scheduled) that detects "platform drift" — where new native Copilot CLI features have made bespoke kopi-cli scaffolding (skills/agents/custom logic) redundant or simplifiable. Inventories the scaffolding, compares each against current native capabilities, classifies KEEP / REMOVE-CANDIDATE / WATCH, and emits a report. Goal: cut scaffolding (Article VII). Every removal needs human approval. Run after a notable Copilot CLI release or before a tech-debt cleanup.