AR
arrive-reviewability
Rules for maintaining reviewable code changes
Install
mkdir -p .claude/skills/arrive-reviewability && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/16747" && unzip -o skill.zip -d .claude/skills/arrive-reviewability && rm skill.zipInstalls to .claude/skills/arrive-reviewability
Activation
This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.
Rules for maintaining reviewable code changes45 charsno explicit “when” trigger
About this skill
<!-- ARRIVE: generated by `arrive sync agent-rules` — do not hand-edit. Edit the canonical source under `arrive/agent-rules/` or `arrive/agent-commands/` and re-run sync. -->
<!-- canonical: arrive/agent-rules/reviewability.md -->
Reviewability Rules
Keep changes bounded and reviewable. TDD/Tidy First sequence: arrive-dev-practices.mdc. Budget thresholds: arrive-core.mdc.
Smaller, Bounded Edits
Prefer smaller, focused changes:
- Isolate tidying - Refactoring commits separate from behavior changes
- Isolate migrations - Database/schema changes in separate PRs
- Isolate auth changes - Security-sensitive code separately
- Separate interface from implementation - Public API changes vs internal refactors
- Split by ARRIVE component - When touching multiple components
Review-Time Usage
- Treat review-time estimate as a planning signal, not a vanity metric.
- Use
arrive scoreoutput to decide checkpointing:- Green: proceed, optional split
- Yellow: split at natural tidy/test/feat boundaries
- Red: split before continuing implementation
- Track estimated and actual review time in the Advance frontmatter and adjust future split strategy when estimates are consistently off.
Evidence Capture
Always run the fastest verification that matches change type:
| Change Type | Required Evidence |
|---|---|
| Logic change | Unit tests (TDD) |
| API change | Integration tests (TDD) |
| UI change | Visual/E2E tests |
| Config change | Lint/validation |
| Tidying | Existing tests still pass |
Attach evidence to the Advance.
Risk Signals
Watch for these risk flags:
resident_touched- Resident component modifiedmigration- Database/schema changesauth- Authentication/authorization changesconcurrency- Threading/async changescaching- Cache invalidation changespublic_api- Public interface changesnew_dependency- New external dependency
Each flag increases reviewability score.