The harness served as one-time review evidence for the tool's attempt
chain; nothing in CI executes bash harnesses and the tool script rarely
changes, so it is removed rather than left to rot (user decision).
- openai searches run an attempt chain: the chat-completions mechanism
(search-preview model + web_search_options patch) first, then the
Responses-native web_search tool on a responses-capable model; the
search fails only when both attempts fail
- both wire-keyed patches (COYOTE_PATCH_OPENAI_CHAT_COMPLETIONS /
COYOTE_PATCH_OPENAI_RESPONSES) are exported per attempt; the client's
resolved wire applies exactly one
- supports_native_search accepts responses-capable openai families
(gpt-4o*/gpt-4.1*/gpt-5*/o3*/o4*)
- stub-coyote harness pins attempt order, patch export, model mapping,
hard failure, and untouched non-openai paths
(scripts/test-web-search-tool.sh)
- CHANGELOG: responses-default breakage caveat replaced with the
both-wire feature note
Black-box scenarios against a built binary with dry_run configs: hook
cadence + env scoping, runtime inertness of failing hooks, --info firing
zero agent.* events, --install-builtins hooks install/force/cold-start
semantics, and end-to-end firing of the installed log-events.sh via the
documented relative-path wiring.
Add a hidden --generate-models flag that rebuilds models.yaml from
models.dev and OpenRouter merged with the hand-owned quirks overlay,
regenerate the committed catalog with it (393 -> 1090 models), and
retire the Python updater. GitHub becomes fully quirks-owned since
upstream only tracks Copilot ids, legacy ids upstream dropped are
preserved as hand-owned quirks entries, and non-chat model families
(image/tts/audio/realtime) are excluded from generation.
Dark-Alex-17