team-rules

# Team Coding Rules Your team keeps its canonical coding rules in a shared Agent Filesystem workspace exposed as the `team-rules` MCP server. The workspace is **read-only** — nobody's agent can write to it. It's the single source of truth for how this team writes code. ## Before writing or editing code 1. Call `mcp__team-rules__file_list` with `path: "/rules"` and `depth: 2`. 2. For the task at hand, read every rule file that plausibly applies. Examples: - auth/secrets/PII work → `/rules/security.md` - any new file or identifier → `/rules/style.md`, `/rules/naming.md` - adding tests → `/rules/testing.md` - cross-module or new module work → `/rules/architecture.md` 3. Cite applicable rules in your plan by path (e.g. "per `/rules/security.md`…"). 4. Apply the rules to the code you produce. ## Rules of engagement - The workspace is read-only. Do not attempt `file_write`, `file_replace`, `file_insert`, or `file_delete_lines` against `team-rules` — they will fail. - If a rule conflicts with what the user asked for, surface the conflict and ask before deciding. - If no rule applies, say so — don't invent or extrapolate. - Prefer reading the live rule over recalling it from memory. Rules change. ## Quick reference | Situation | Rule file to read | |-----------|------------------| | Any code edit | `/rules/style.md` | | New file / function / type | `/rules/naming.md` | | Auth, input handling, secrets, PII | `/rules/security.md` | | Adding or modifying tests | `/rules/testing.md` | | New module or cross-boundary work | `/rules/architecture.md` | | Commit messages, PR description | `/rules/git.md` |