Skip to content

code_review.md

Latest commit

 

History

History
45 lines (34 loc) · 1.68 KB

code_review.md

File metadata and controls

45 lines (34 loc) · 1.68 KB

Code Review Guide

Use this checklist when reviewing local changes, PRs, commits, or agent output. Lead with concrete findings and file references.

Review Priorities

  • Correctness: Does the change satisfy the requested behavior?
  • Regression risk: Could existing workflows, schemas, CLI contracts, or runtime state be broken?
  • Verification: Are tests, lint, build, or honest substitutes run and reported?
  • Scope control: Are unrelated refactors, generated files, caches, or local runtime artifacts excluded?
  • Security and privacy: Are secrets, transcripts, credentials, screenshots, or sensitive local paths avoided?
  • Observability: For AgentLens/Waygent changes, are durable artifacts, event schemas, and non-blocking behavior preserved?

AgentLens Checks

  • JSON artifacts remain the source of truth; SQLite stays a rebuildable cache.
  • AgentLens internal failures must not change the wrapped command exit code.
  • Schema changes are additive unless a versioned migration is explicitly designed.
  • Dashboard/API type drift is checked with npm run gen-types when relevant.

Waygent Runtime Checks

  • Providers do not write SQLite or AgentLens directly.
  • Scheduler changes respect safe waves, dependency checkpoints, and failure barriers.
  • Recovery paths stop on missing handlers or human-decision classes instead of recording fake progress.
  • Runtime behavior changes include targeted tests or scenario harness coverage.

Output Format

For review responses:

  1. Findings first, ordered by severity.
  2. File and line references when possible.
  3. Open questions or assumptions.
  4. Short verification summary.

If there are no findings, say that directly and mention any residual risk.