Repository navigation
[cli-tools-test] audit MCP request times out after 60 seconds without partial results or recovery guidance #66029
Description
Problem Description
A recent-run audit request through the agentic-workflows MCP interface terminated after 60.0 seconds with context deadline exceeded. It returned no audit report, partial results, or actionable recovery guidance. This blocks the logs-to-audit investigation flow.
Command/Tool
- Tool:
audit - MCP arguments:
{"run_id":"37415703249","max_tokens":3000} - CLI transport:
agenticworkflows audit '{"run_id":"37415703249","max_tokens":3000}'
Steps to Reproduce
- In the configured agentic-workflows MCP environment for
github/gh-aw, requestlogswith{"engine":"copilot","start_date":"-7d","count":3,"max_tokens":3000}. - Observe the returned continuation cursor containing
before_run_id: 37415703249. - Call
auditwith the arguments above. The run identifier is a cursor value, not independently verified as one of the three downloaded runs. - Observe the request timing out without an audit payload.
Expected Behavior
The audit should complete, or return an actionable timeout response explaining the incomplete analysis and any supported recovery. A caller should be able to distinguish a run-specific data problem from a request deadline.
Actual Behavior
The tool returned only Error [0]: context deadline exceeded after 60,011 ms (60,041 ms total bridge time). No partial report was returned.
A separate audit of §37418345812 completed in 58.1 seconds and included in-progress metadata, job timelines, tool metrics, MCP health, and network analysis. Both observations exceed this exploratory test's 30-second audit target, but do not establish a general latency regression.
Environment
- Repository: github/gh-aw
- Testing run: §37418327052
- Date: 2026-10-06 UTC
- gh-aw version: Not supplied by
status; no directgh awinvocation was used.
Impact
- Severity: Medium
- Frequency: Observed once; not retried. Reproducibility and root cause remain unconfirmed.
- Workaround: None validated.
Logs/Diagnostics
View timing and error diagnostics
MCP tools/call: status=200, elapsed=60011ms
Tool call complete: total=60041ms
Error [0]: context deadline exceeded
Tool call error: Error [0]: context deadline exceeded
The repository's pkg/cli/mcp_tools_privileged.go defines defaultMCPAuditTimeoutMinutes = 5 and uses it when creating the audit subprocess context. This does not identify which layer caused the observed 60-second request failure. No infrastructure inspection or timeout modification was attempted.
max_tokens is documented in the source schema as deprecated and ignored; it is included here to reproduce the actual request, not as a proposed fix.
Additional Context
Issue duplicate searches were attempted through the GitHub read tool, but results were removed by integrity policy. Existing duplicates could not be assessed. Please check for an existing timeout issue before triage.
Next step: investigate deadline/error propagation for this audit request using approved diagnostics, preserving configured limits.
References:
Generated by 🧪 Daily Cli Tools Tester · codex · gpt60 · 187.9 AIC · ⌖ 36.8 AIC · ⊞ 19.6K · ◷
- expires on Oct 12, 2026, 9:32 PM UTC-08:00
Activity