coder/AGENTS.md

Coder Development Guidelines

Make the smallest correct change, follow existing patterns, and verify the result. Carry the requested task through implementation, verification, and necessary follow-up until it is complete or blocked by information or access you cannot obtain. Do not stop at a plan, partial fix, or offer to continue when the user requested completed work.

Prioritize correctness over agreement. State uncertainty instead of guessing, and push back on technically unsound requests with evidence.

Autonomy and clarification

• Resolve routine ambiguity by inspecting relevant code, tests, documentation, and history. Make reasonable, reversible assumptions consistent with the user's intent and existing patterns; state consequential assumptions and continue working.
• Ask only when essential information cannot be recovered from available context and would materially change the result, or when a destructive or irreversible action requires authorization the user has not already provided. Reuse authorization from the conversation instead of asking again for the same action.
• If clarification or approval is required, continue authorized work that does not depend on the answer. Explain the specific blocker and what you have already investigated.
• Apply repository guidance within its stated scope and honor explicit user instructions. Do not turn optional recommendations or routine implementation choices into approval requirements.

Task-specific guidance

Load only the guidance relevant to the task:

Scope	Guidance
Development servers, Git, hooks, and routine checks	WORKFLOWS.md
API endpoints and Swagger	WORKFLOWS.md
Go	GO.md
Tests and concurrency	TESTING.md
Database and SQLC	DATABASE.md
OAuth2 and authorization	OAUTH2.md
Architecture	ARCHITECTURE.md
Troubleshooting	TROUBLESHOOTING.md
Observability	OBSERVABILITY.md
Isolation, ports, and cleanup	DEV_ISOLATION.md
Failure reports	AGENT_FAILURES.md
PR descriptions	PR_STYLE_GUIDE.md
Existing docs prose	docs style guide
Docs scope and routing	content guidelines
Docs structure and research	DOCS_STYLE_GUIDE.md
New, moved, or restructured docs	write-docs skill
Frontend	site/AGENTS.md

For changes under site/src/, also read FRONTEND_PATTERNS.md. When you modify frontend code under site/, run the frontend-deep-review skill and fix every blocking and should-fix finding before handoff. For chatd work, read coderd/x/chatd/ARCHITECTURE.md. When the docs style guide and the content guidelines conflict, the content guidelines govern scope and routing. Automated tooling checks a small subset of the style guide and Vale runs advisory, so read the guide and apply it; a clean make lint/prose is not conformance. What the tooling checks, and what it doesn't has the current numbers.

Workflow

• Inspect the working tree before editing. For an existing PR, check out its branch first.
• Follow existing architecture for routine decisions. For requested architectural work, investigate options, choose a reasonable approach, and explain the tradeoffs while proceeding. Ask before introducing major architectural changes outside the requested scope.
• Answer informational questions directly. Requests to implement, fix, or investigate authorize that work even when phrased as a question.
• Install and use the repository Git hooks. Never bypass them with --no-verify. Wait for slow first runs while caches warm.
• Prefer targeted tests and checks while iterating. Run the broader checks required by the affected area before handoff.
• Do not force-push unless explicitly requested.
• Commit and PR titles use type(scope): message. A scope must be a real path containing every changed file. Use a broader scope or no scope for cross-cutting changes.
• Name branches so they do not collide with issue-tracker IDs. When a branch references a GitHub issue number, write it as issue-<number> (for example, issue-1234-fix-flake), not as a bare <word>-<number> such as docs-1234 or a leading number. Connected trackers like Linear auto-link any branch containing a <key>-<number> token to the same-numbered issue on the team that owns <key>, silently attaching the PR to an unrelated ticket and moving it through that ticket's workflow.

Essential commands

Task	Command
Develop	./scripts/develop.sh
Build	make build
Build slim	make build-slim
Test	make test
Test one	make test RUN=TestName
Race test	make test-race
Lint	make lint
Generate	make gen
Format	make fmt
Pre-commit checks	make pre-commit
Pre-push checks	make pre-push

Docs use pnpm run format-docs and pnpm run lint-docs. Frontend commands live in site/AGENTS.md.

Repository guardrails

• Database changes: edit coderd/database/queries/*.sql, run make gen, update enterprise/audit/table.go for audit errors, then run make gen again.
• New resources: scope every new resource to an organization (organization_id column, organization-scoped RBAC and routes), never deployment-wide.
• OAuth2: return RFC-compliant errors with httpapi.WriteOAuth2Error(...). Public endpoints that need system access use dbauthz.AsSystemRestricted.
• Chatd: when a change affects the documented architecture, do not edit the architecture document yourself. Leave TODO items in the affected sections; the human PR author writes the actual updates.
  • When you review a PR, treat changes to that document as the owner's work. TODOs are notes for the author, so flag any that the PR adds to the architecture document.
• Public API: add the required Swagger annotations for new public HTTP endpoints.
• Transactions: keep InTx work on the transaction handle. Prefer explicit database-to-SDK converters.
• Concurrent tests: call t.Parallel(), use unique identifiers, and do not use time.Sleep to mask timing problems.
• Frontend: reuse shared UI primitives. Prefer Vitest and userEvent tests that assert the non-visual outcome of the interaction (callback, request, state); extend existing coverage instead of adding a new test when equivalent coverage already exists. Use Storybook stories only for visual components that should be covered by the visual regression tool, which Pixel screenshots in CI.
• GitHub Actions: set top-level permissions: {} and grant only required permissions per job.

Code and writing style

• Follow the Uber Go Style Guide for Go code.
• Use language-server navigation when available.
• Name code for what it does, not its implementation or history. Wrap errors with context.
• Document exported symbols with idiomatic Go doc comments or JSDoc.
• Avoid unrelated edits. Preserve comments that explain non-obvious behavior.
• Comments must be concise and substantive. Explain behavior, constraints, or rationale, not the history of the edit.
• Do not use em dashes, en dashes, or spaced double hyphens as punctuation in code, comments, strings, or documentation.
• Ensure files end with a newline.

Local configuration

Read AGENTS.local.md when present. It may be gitignored and is not imported automatically.