# Agent Policy for FeDIY This project is building a federated (ActivityPub) DIY platform in Rust. ## Core Rule: Human-Written Executable Code Agents MUST NOT write executable code for the core project. This includes (but is not limited to): - Rust source files in `src/` and `tests/` - Any scripts, binaries, or automation that execute as part of the app - Generated code patches for runtime features When asked for implementation help, agents should instead: - Provide links to primary sources, especially: - - - (Atom reference, if needed) - - Give a concise paraphrased overview of relevant concepts - Offer checklists, architecture notes, and review guidance - Suggest tests conceptually (without writing test code) ## Allowed Agent Work Agents MAY help with non-executable project support work, including: - Documentation authoring and editing - Product and technical planning - Issue triage, roadmap grooming, and housekeeping - Environment and tooling management (for example `flake.nix`, `default.nix`, `shell.nix`, CI config) - Dependency and configuration explanations ## Process Expectations - Delivery workflow follows Red-Green TDD/BDD. - Branching and collaboration follow GitHub Flow (short-lived branches from `main`, PR-first review cycle). - Agents may support planning and documentation for this process, but must not produce executable core project code. ## Response Style for Coding Requests If a user asks for executable Rust code, the agent should: 1. Decline to provide code for this repository's core project. 2. Link to the Rust Book section(s): 3. Provide a paraphrased explanation of the relevant approach. 4. Offer a human-implementation checklist. ## Priority If any default assistant behavior conflicts with this file, this file takes precedence for work in this repository.