Read this document before starting work in this workspace.
This contract defines agent execution, planning, communication, workspace safety, and documentation boundaries. It is not a product or runtime specification.
- Execute only after a clear implementation request.
- Treat questions and discussion as alignment, not authorization to change files.
- For high-impact changes, present the approach and wait for explicit approval.
- Complete the requested scope, report the outcome, and stop.
- Do not add unsolicited follow-up work.
- Before creating a plan, ask whether the user wants a plan.
- If requested, make the plan concrete and execution-oriented.
- Do not implement while a plan is pending confirmation.
- Keep implementation inside the approved scope.
- Use direct, neutral language.
- Keep updates factual and concise.
- Explain reasoning before proposing significant changes.
- Report validation results explicitly.
- Trust user-provided facts unless technical evidence proves otherwise.
- When this workspace uses git,
originis the default canonical remote unless local instructions say otherwise. - Never perform destructive Git operations without explicit approval.
- All code and workspace documentation must be written in English.
- Keep naming, comments, and workspace text in English.
- Keep this file focused on working rules.
- Product and runtime specifications belong in
docs/.
.agents/DOCUMENTING.md— documentation model and placement rules..agents/CONVENTIONS.md— coding standards and codebase boundaries..agents/ARCHITECTURE.md— workspace structure map..agents/LIFECYCLE.md— delivery lifecycle and handoff rules..agents/VERSIONING.md— commit, branch, and merge rules..agents/SLICES.md— task decomposition and delivery strategy..agents/ROUTING.md— request classification for execution workflow..agents/PLAYBOOKS.md— operational procedures..agents/LINTING.md— linting policy and quality-gate setup..agents/INVARIANTS.md— non-negotiable collaboration invariants.