Skip to content

fix: improve issue-triage prompt robustness to prevent no-output failures#10504

Merged
timotheeguerin merged 2 commits into
mainfrom
copilot/debug-agentic-triage-failure
Apr 27, 2026
Merged

fix: improve issue-triage prompt robustness to prevent no-output failures#10504
timotheeguerin merged 2 commits into
mainfrom
copilot/debug-agentic-triage-failure

Conversation

Copy link
Copy Markdown
Contributor

Copilot AI commented Apr 27, 2026

Agentic Triage workflow (run 24996553319) produced no safe outputs despite the agent running for ~4 minutes. This is the second failure after PR #10502 fixed tool name references — the agent still silently failed to call any safe output tool.

Analysis of the generated prompt and MCP gateway logs revealed three issues:

  • Missing get_file_contents tool reference: The prompt said "read the file eng/common/config/labels.ts" without specifying the MCP tool name or parameters. Now explicitly references get_file_contents with owner/repo.
  • Unconditional web-fetch reference: The prompt unconditionally instructed the agent to "Use web-fetch to verify the playground link", but only safeoutputs and github MCP servers were registered — web-fetch was not available. Changed to conditional: "If the web_fetch tool is available, use it to verify..."
  • No fallback instruction: Added an explicit ## Important section reinforcing the requirement to call at least one safe output tool (add_comment, add_labels, or noop) before finishing, with guidance to call noop on errors.

Copilot AI linked an issue Apr 27, 2026 that may be closed by this pull request
…ures

Three targeted fixes to the Agentic Triage workflow prompt:

1. Add explicit `get_file_contents` tool reference with owner/repo params
   for reading the labels configuration file, so the agent knows which
   MCP tool to use.

2. Make web-fetch playground verification conditional ("if the tool is
   available") since web_fetch may not be registered as an MCP tool in
   all environments.

3. Add a fallback "Important" section reinforcing the requirement to call
   at least one safe output tool before finishing, with explicit guidance
   to call `noop` if errors are encountered.

Agent-Logs-Url: https://github.com/microsoft/typespec/sessions/1a40bdca-9a2a-4fae-bf5f-42d3cc86847c

Co-authored-by: timotheeguerin <1031227+timotheeguerin@users.noreply.github.com>
Copilot AI changed the title [WIP] Debug Agentic Triage workflow failure fix: improve issue-triage prompt robustness to prevent no-output failures Apr 27, 2026
Copilot AI requested a review from timotheeguerin April 27, 2026 13:40
@timotheeguerin timotheeguerin marked this pull request as ready for review April 27, 2026 13:45
@timotheeguerin timotheeguerin merged commit cd27a87 into main Apr 27, 2026
24 of 27 checks passed
@timotheeguerin timotheeguerin deleted the copilot/debug-agentic-triage-failure branch April 27, 2026 13:45
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

Projects

None yet

Development

Successfully merging this pull request may close these issues.

[aw] Agentic Triage failed

2 participants