Skip to content

Re-diagnose q workflow success rate as command gating, not execution failure - #54854

Merged
pelikhan merged 3 commits into
mainfrom
copilot/fix-9919-1036865607-f7a577f4-ad52-44b9-8f08-81d2ea488992
Aug 22, 2026
Merged

Re-diagnose q workflow success rate as command gating, not execution failure#54854
pelikhan merged 3 commits into
mainfrom
copilot/fix-9919-1036865607-f7a577f4-ad52-44b9-8f08-81d2ea488992

Conversation

CopilotAI commented Aug 22, 2026

Copy link
Copy Markdown
Contributor

The reported "0.8% execution success" for the q workflow is not a failure mode. q is a slash-command workflow, so GitHub starts a run for every issue/comment/discussion event and the generated gate stops it unless the body starts with /q — 2 activated runs out of 261 at ~9.9s average is the designed behavior. PR #43527 (shared prompt quality gate for agent-review workflows) was never wired into q; that attribution was wrong from the start and kept being carried forward through shared-alerts.md.

The gate is visible in the compiled activation condition:

jobs:
activation:
needs: pre_activationif: "needs.pre_activation.outputs.activated == 'true' && (... startsWith(github.event.comment.body, '/q ') ...)"

q itself needs no change. The defect is in how the meta-orchestrators score and remember it.

agent-performance-analyzer.md

  • New Command- and mention-triggered workflows rule: for slash_command / command / mention workflows, score only activated runs and report the rest as "command gating (expected)" instead of as execution-success or activation-refused problems.
  • New Root-cause hygiene rule: re-verify any root cause carried over from shared-alerts.md / workflow-health-latest.md; if it names a PR/issue as the pending fix and that PR/issue is merged or closed while the metric is unchanged, drop it and re-diagnose from trigger config and run logs.
  • Mirrored the gating rule in the detailed Phase 1 run-analysis step, and added a pruning bullet to the shared-alerts.md write step.

workflow-health-manager.md

  • Same two rules in the execution-monitoring section, plus the shared-alerts.md pruning bullet.

Lock files

  • Regenerated agent-performance-analyzer.lock.yml and workflow-health-manager.lock.yml (prompt body is runtime-imported, so only the body hash changes).

Note: shared-alerts.md lives in runtime repo-memory on the memory/meta-orchestrators branch and can't be edited from here — the pruning rule is what makes the next orchestrator run remove the stale note.

CopilotAIand others added 2 commits August 22, 2026 16:21
Co-authored-by: pelikhan <4175913+pelikhan@users.noreply.github.com>
…ification in meta-orchestrators
Co-authored-by: pelikhan <4175913+pelikhan@users.noreply.github.com>
CopilotAI changed the title [WIP] Copilot RequestRe-diagnose q workflow success rate as command gating, not execution failureAug 22, 2026
CopilotAI requested a review from pelikhanAugust 22, 2026 16:25
@pelikhan
pelikhan marked this pull request as ready for review August 22, 2026 16:45
CopilotAI balanced review requested due to automatic review settings August 22, 2026 16:45
@pelikhan
pelikhan merged commit a5536e5 into mainAug 22, 2026
1 check passed
@pelikhan
pelikhan deleted the copilot/fix-9919-1036865607-f7a577f4-ad52-44b9-8f08-81d2ea488992 branch August 22, 2026 16:45

CopilotAI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Pull request overview

Reclassifies q’s low activation ratio as expected slash-command gating and prevents stale attribution to PR #43527.

Changes:

  • Updates analyzer and health-manager prompts.
  • Regenerates their body hashes.
  • Includes unrelated generated workflow changes, including strict-mode downgrades.
Show a summary per file
FileDescription
agent-performance-analyzer.mdAdds gating and root-cause guidance.
agent-performance-analyzer.lock.ymlUpdates prompt body hash.
workflow-health-manager.mdMirrors diagnostic guidance.
workflow-health-manager.lock.ymlUpdates prompt body hash.
smoke-drive.lock.ymlUnrelated shell regeneration.
mcp-inspector.lock.ymlDisables compiled strict mode.
gpclean.lock.ymlDisables compiled strict mode.
example-permissions-warning.lock.ymlDisables compiled strict mode.
dev.lock.ymlDisables compiled strict mode.
daily-team-evolution-insights.lock.ymlDisables compiled strict mode.
copilot-pr-merged-report.lock.ymlDisables compiled strict mode.
cli-version-checker.lock.ymlDisables compiled strict mode.
cli-consistency-checker.lock.ymlDisables compiled strict mode.
blog-auditor.lock.ymlDisables compiled strict mode.
ace-editor.lock.ymlDisables compiled strict mode.

Review details

  • Files reviewed: 15/15 changed files
  • Comments generated: 15
  • Review effort level: Balanced

Comment on lines +117 to +118
**Command- and mention-triggered workflows:** Workflows whose frontmatter uses `slash_command`,
`command`, or `mention` (for example `q`) are wired to `issues`, `issue_comment`, `discussion`, and
Comment on lines +126 to +131
**Root-cause hygiene:** Before repeating a root cause carried over from shared memory
(`shared-alerts.md`, `workflow-health-latest.md`), re-verify it against current evidence. If the note
cites a PR or issue as the pending fix, check its current state: when it is already merged or closed
and the metric has not moved, the attribution is stale — delete or replace it in `shared-alerts.md`
and re-diagnose from the workflow's trigger configuration and recent run logs instead of restating
it.
Comment on lines +121 to +122
- Exclude command- and mention-triggered workflows (frontmatter `slash_command`, `command`, or
`mention`, for example `q`) from execution-rate scoring. They start a run for every issue,
Comment on lines +127 to +130
- Re-verify any root cause carried over from `shared-alerts.md` or `workflow-health-latest.md` before
repeating it. If the note names a PR or issue as the pending fix and that PR/issue is already
merged or closed while the metric is unchanged, the attribution is stale: remove or replace it in
`shared-alerts.md` and re-diagnose from the workflow's trigger configuration and recent run logs.
GH_AW_INFO_CACHE_MEMORY: "true"
GH_AW_INFO_FRONTMATTER_EMOJI: "🔍"
GH_AW_COMPILED_STRICT: "true"
GH_AW_COMPILED_STRICT: "false"
GH_AW_INFO_CACHE_MEMORY: "true"
GH_AW_INFO_FRONTMATTER_EMOJI: "🔢"
GH_AW_COMPILED_STRICT: "true"
GH_AW_COMPILED_STRICT: "false"
GH_AW_INFO_AGENT_RUNTIME: "cloud-hypervisor"
GH_AW_INFO_FRONTMATTER_EMOJI: "✅"
GH_AW_COMPILED_STRICT: "true"
GH_AW_COMPILED_STRICT: "false"
GH_AW_INFO_AGENT_RUNTIME: "cloud-hypervisor"
GH_AW_INFO_FRONTMATTER_EMOJI: "📝"
GH_AW_COMPILED_STRICT: "true"
GH_AW_COMPILED_STRICT: "false"
GH_AW_INFO_AGENT_RUNTIME: "cloud-hypervisor"
GH_AW_INFO_FRONTMATTER_EMOJI: "✏️"
GH_AW_COMPILED_STRICT: "true"
GH_AW_COMPILED_STRICT: "false"
Comment on lines +1810 to +1811
if [ -d /tmp/gh-aw/sandbox/firewall/logs ]; then mkdir -p /tmp/gh-aw/threat-detection/sandbox/firewall/logs && cp -r /tmp/gh-aw/sandbox/firewall/logs/. /tmp/gh-aw/threat-detection/sandbox/firewall/logs/; fi
if [ -d /tmp/gh-aw/sandbox/firewall/audit ]; then mkdir -p /tmp/gh-aw/threat-detection/sandbox/firewall/audit && cp -r /tmp/gh-aw/sandbox/firewall/audit/. /tmp/gh-aw/threat-detection/sandbox/firewall/audit/; fi
Sign up for freeto join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

q workflow: persistent 0.8% execution success needs re-diagnosis (stale root cause)

3 participants

@pelikhan