Zamil Majdy 98f13a6e5d feat(copilot): add create -> dry-run -> fix loop to agent generation (#12578) ...

## Summary
- Instructs the copilot LLM to automatically dry-run agents after
creating or editing them, inspect the output for wiring/data-flow
issues, and fix iteratively before presenting the agent as ready to the
user
- Updates tool descriptions (run_agent, get_agent_building_guide),
prompting supplement, and agent generation guide with clear workflow
instructions and error pattern guidance
- Adds Tool Discovery Priority to shared tool notes (find_block ->
run_mcp_tool -> SendAuthenticatedWebRequestBlock -> manual API)
- Adds 37 tests: prompt regression tests + functional tests (tool schema
validation, Pydantic model, guide workflow ordering)
- **Frontend**: Fixes host-scoped credential UX — replaces duplicate
credentials for the same host instead of stacking them, wires up delete
functionality with confirmation modal, updates button text contextually
("Update headers" vs "Add headers")

## Test plan
- [x] All 37 `dry_run_loop_test.py` tests pass (prompt content, tool
schemas, Pydantic model, guide ordering)
- [x] Existing `tool_schema_test.py` passes (110 tests including
character budget gate)
- [x] Ruff lint and format pass
- [x] Pyright type checking passes
- [x] Frontend: `pnpm lint`, `pnpm types` pass
- [x] Manual verification: confirm copilot follows the create -> dry-run
-> fix workflow when asked to build an agent
- [x] Manual verification: confirm host-scoped credentials replace
instead of duplicate

2026-04-03 14:48:57 +00:00

0 lines

0 B

Python

Raw Blame History

The file is empty.