| name | tooltest-fix-loop |
| description | Use when running tooltest to validate MCP servers, interpret failures, and iterate fixes in this repo. |
Tooltest fix loop
Run tooltest and iterate
- Identify how to start the repo’s MCP server (stdio or streamable HTTP).
- Run tooltest against it:
tooltest stdio --command "<server start command>"tooltest http --url "<server mcp url>"
- Read the failure report, trace the failing tool call or protocol step, and fix the smallest viable patch.
- Re-run tooltest until it exits 0; keep tool names and schemas stable unless a spec violation requires change.
Triage guidance
- Treat failures as either protocol handling, tool schema validation, or runtime behavior; fix in that order.
- Prefer backward-compatible fixes; add a regression test when a failure is deterministic or easy to reproduce.
- When the failure is intermittent, keep the smallest reproduction from the report and turn it into a test.
Reporting
- Summarize the changes and why they address the specific tooltest failures.
- Include the final passing tooltest output snippet.