Copilot Role
GitHub Copilot is primarily used for:
- GitHub-native PR review
- local coding assistance
- focused implementation suggestions
- small bug and test-gap detection
- concise explanations of changed code
Keep
policy (binding)
The 90% line-coverage gate is the floor, not the target.
- Adding code drops coverage? Add tests in the SAME change set to land back at ≥ 90%. Never
first user.message starts with "Researching: "
isReview: boolean // true if events.jsonl contains "agent_type":"code-review"
agentTypes: string[] // all agent_type values found in events.jsonl (e.g. ["explore", "security-reviewer"])
researchReports: string
designed for AI agent consumption (AgentSkills standard)
- **AI Agent**: Tools like GitHub Copilot, Claude Code, or similar assistants
- **Trunk-Based Development**: All changes committed directly to main branch
## Quality Standards
Review Required**: All pull requests require human approval. Treat agent contributions like code from a junior developer—review thoroughly, request changes, and iterate.
## Pull Request Guidelines
- Keep PRs focused
owning skills, for example `php-modernization`, `typo3-content-blocks`, and `typo3-update`.
- Always review AI-generated code before committing.
- When multiple skills are relevant, combine them, for example `typo3-rector
cleanup, revert, or apply approval.
- Do not delete docs without replacement or modify .env/secrets.
- Coding agents never run `git commit` or `git push`; the user performs both manually.
- Forwarded
versioning
## Development Guidelines
- **Complete implementations**: No TODOs, placeholders, or missing pieces
- **Readability over performance**: Code should be clear and maintainable
- **Add tests for bugs**: When fixing bugs, add test cases
Architecture & structural soundness" sections
apply to **both** writers (follow them when generating new code) and
reviewers (see [Reviewer etiquette](#reviewer-etiquette-nits--architecture)
for when to surface them).
## PR Review
Never bulk-load it. Plans and reviews there are evidence of past reasoning, not a description of the product today.
- **On conflict, the code wins for behavior; active docs
SwiftUI code without mobile-a11y-specialist review
4. Do NOT write concurrent code without concurrency-specialist review
5. Do NOT handle credentials or encryption without swift-security-specialist review
Review Sidecar Format), also known as **Sidemark**, is a specification plus tooling for storing review comments in sidecar files alongside Markdown documents. The canonical spec is `MRSF-v1.0.md`.
## Repository Architecture
This
integration of the entire branch into main**, so describe what's being integrated.
## CodeReview Requirement
**After any changes to files under `src/` or `tests/`**, a codereview MUST
starting work, check your capability profile in `.squad/team.md` under the **Coding Agent → Capabilities** section.
- **🟢 Good fit** — proceed autonomously.
- **🟡 Needs review** — proceed, but note in the PR description that a squad
verify the plan with: "Are you happy with this implementation plan?" before proceeding with code changes.
6. Reference related breadcrumbs when a task builds on previous work.
7. Before concluding