Coding (Implementation + Review)

Use this skill when the primary intent is implementation, bug fixing, refactoring, code-quality review, or technical guidance for repository code changes.

Scope and routing

• Use this skill for code edits, code-quality reviews, PR feedback loops, and infra/platform implementation tasks.
• If the request is planning/spec/lifecycle management, hand off to planning.
• If the request is AGENTS architecture housekeeping, hand off to housekeeping.
• If the request is under skills/ or edits SKILL.md, hand off to skill-creator.
• For frontend framework guidance, use SolidJS references only (skills/coding/references/solidjs/...).

Mandatory smell baseline (always for code changes)

For implementation/bugfix/refactor/review operations, always open:

• skills/coding/references/code-smells/index.md
• skills/coding/references/code-smells/detection-signals.md

Then open specific smell files for detailed refactor guidance when detection signals match. When writing new code (not just editing), err on the side of reading more smell refs — it is cheaper to check and dismiss than to miss a smell.

Core workflow

1. Confirm scope, constraints, and acceptance criteria.
2. Load mandatory smell baseline refs.
3. Load domain refs relevant to the task.
4. Read only necessary code paths.
5. Implement minimal focused changes.
6. Self-review: check your own output against detection signals. Fix violations before presenting — especially Duplicate Code, Speculative Generality, and Long Method, which are easy to introduce organically.
7. Validate with targeted checks/tests.
8. Summarize changes, validations, and remaining risks.

Domain reference triggers (open when clearly relevant)

• Code smell / maintainability / quality diagnostics:
  • skills/coding/references/code-smells/index.md
  • skills/coding/references/code-smells/smells/index.md
• Work-package refactoring execution:
  • skills/coding/references/refactoring/index.md
• SolidJS implementation/performance:
  • skills/coding/references/solidjs/index.md
  • skills/coding/references/solidjs/rules/index.md
• Infra / deploy / platform ops:
  • skills/coding/references/platform-engineering/index.md
• Auth / credentials / secret handling:
  • skills/coding/references/secrets-and-auth-guardrails.md
• PR review bot loop / CI failure remediation:
  • skills/coding/references/gh-pr-review-fix.md
• JS/TS runtime and toolchain:
  • skills/coding/references/bun.md

Quality rules

• Prefer hard cutovers over fallback-first compatibility branches.
• Avoid over-defensive code that obscures normal control flow.
• Preserve existing architecture unless the task explicitly asks for structural change.
• Add/update tests when behavior changes.
• Keep edits small, cohesive, and traceable to the request.

Output expectations

• Explain decisions and key trade-offs.
• List changed files.
• Report validations run (or why skipped).
• Call out unresolved risks and follow-up actions.