diff --git a/FULLSEND.md b/FULLSEND.md new file mode 100644 index 0000000..02eba67 --- /dev/null +++ b/FULLSEND.md @@ -0,0 +1,46 @@ +# Using fullsend with this repository + +This repository uses [fullsend](https://github.com/fullsend-ai/fullsend) +for AI-assisted development. Fullsend runs as a set of agent roles that +respond to GitHub events (issues, pull requests, comments) via the +`.github/workflows/fullsend.yaml` workflow. + +## Configured roles + +The `.fullsend/config.yaml` file enables six roles: + +| Role | Trigger | What it does | +|------|---------|--------------| +| **triage** | Issue opened, edited, or labeled | Analyzes the issue, identifies root cause, and proposes a plan | +| **coder** | `/fs-code` comment on a triaged issue | Implements the fix or feature and opens a pull request | +| **review** | Pull request opened or updated | Reviews the PR for correctness, style, and completeness | +| **fix** | Review requests changes | Automatically addresses review feedback on agent PRs | +| **retro** | Pull request merged | Runs a retrospective to capture lessons learned | +| **prioritize** | On demand | Assesses and prioritizes open issues | + +## How to interact + +1. **Open an issue** describing a bug or feature request. The + **triage** agent analyzes it and adds a `ready-to-code` label when + the issue is actionable. +2. **Trigger the coder** by commenting `/fs-code` on a triaged issue. + The **coder** agent creates a branch, implements the change, and + opens a PR. +3. **Review happens automatically.** The **review** agent examines the + PR and posts findings. If changes are requested, the **fix** agent + attempts to resolve them. +4. **Merge the PR** once it passes review. The **retro** agent runs a + retrospective after merge. + +## Stopping the fix agent + +If you want to prevent the fix agent from acting on a PR, comment +`/fs-fix-stop`. This adds the `fullsend-no-fix` label and disables +automatic fixes for that PR. Remove the label or comment `/fs-fix` +to re-engage. + +## Configuration + +The fullsend configuration lives in `.fullsend/config.yaml`. See the +[fullsend documentation](https://github.com/fullsend-ai/fullsend) for +available options.