feat: distribute agent detection manifests

1 files changed, 7 insertions, 1 deletions

diff --git a/AGENTS.md b/AGENTS.md
index 7658a051..f04fbed5 100644
--- a/AGENTS.md
+++ b/AGENTS.md
@@ -9,7 +9,7 @@ Terminal workspace manager for AI coding agents. Rust + ratatui.
 - **No god objects.** If a module is doing too many things, split it. `app/` is already split into state, actions, and input. Keep it that way.
 - **Platform code is isolated.** OS-specific behavior lives in `src/platform/`. Core modules don't have `#[cfg(target_os)]`.
 - **Detection is decoupled.** The detector reads a screen snapshot, never touches the parser or viewport state.
-- **Screen detection is evidence-based.** When changing `src/detect/agents/`, first capture the relevant bottom-buffer state with `herdr pane read --source recent --format text` and, when styling or alternate screen behavior matters, `--format ansi`. Decide which visible controls are invariant, which are alternatives, and encode them as explicit AND/OR gates. Do not match whole-pane incidental text, and do not use the user-visible viewport for agent status because users can scroll it.
+- **Screen detection is evidence-based.** When changing `src/detect/manifests/`, first capture the relevant bottom-buffer state with `herdr agent read <pane> --source detection --format text` and, when styling or alternate screen behavior matters, `--format ansi`. Decide which visible controls are invariant, which are alternatives, and encode them as explicit AND/OR gates. Do not match whole-pane incidental text, and do not use the user-visible viewport for agent status because users can scroll it.
 - **UI patterns should be reused.** Herdr is a mouse-first TUI. New dialogs, onboarding, settings, and post-update flows should follow the existing UI/UX language and interaction patterns instead of inventing one-off screens. Prefer reusing existing modal/screen structure, affordances, and close actions so the app feels consistent.
 
 ## Multi-agent isolation
@@ -49,6 +49,12 @@ Run `just check` before committing unless Can explicitly accepts narrower valida
 
 Unit tests live next to the code (`#[cfg(test)] mod tests`). New `AppState` or `Workspace` behavior should be testable with `AppState::test_new()` and `Workspace::test_new()` without PTYs.
 
+## Agent Detection Updates
+
+Agent detection changes should use the manifest hot-reload loop. Can drives the real agent UI into the target state, then you read the pane with `herdr agent read <pane> --source detection --format text` and inspect matching with `herdr agent explain <pane> --json`. Update the bundled manifest in `src/detect/manifests/<agent>.toml`, copy that manifest to the local override path at `~/.config/herdr/agent-detection/<agent>.toml`, then run `herdr server reload-agent-manifests`. Can verifies the live pane state, and once the rule is correct, remove the local override so the committed bundled manifest remains the source of truth.
+
+Do not add large agent-specific full-screen fixture suites for routine manifest tuning. Keep Rust tests focused on manifest parsing, rule semantics, skip-state semantics, source precedence, cache reload behavior, and update flow. Use live pane reads for agent-specific screen evidence.
+
 ## Vendored libghostty-vt
 
 `vendor/libghostty-vt.vendor.json` records the upstream source commit currently vendored.