You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
threat detection engine error
The threat detection engine encountered an error and could not complete analysis. This is a tooling failure, not a security finding.
Details
The threat detection engine failed to produce results.
The compile tool in the agentic-workflows MCP server appears to default strict to true, even when the --strict flag is not passed. This causes the generated .lock.yml to differ from the committed version for workflows that were originally compiled without strict mode (e.g. ace-editor.md), producing unwanted diffs and dropping the Enforce strict mode policy step.
git status shows a clean working tree for .github/workflows/.
Run agenticworkflows compile with {"workflows":["ace-editor"]} (no strict param supplied).
Run git diff .github/workflows/ace-editor.lock.yml.
Expected Behavior
Without an explicit strict parameter, compile should use the workflow's existing/previous strict setting (or default false, matching gh aw compile CLI default), leaving the lock file unchanged when the markdown source hasn't changed.
and the gh-aw-metadata header gains a "strict":true field. It also removes the Enforce strict mode policy step for the previous (non-strict) mode. Simply running compile via the MCP tool without special flags mutates lock files that would otherwise be byte-identical, and could mask/alter workflow policy enforcement behavior.
Environment
Repository: github/gh-aw
Run ID: 30687098351
Date: 2026-08-01
Impact
Severity: Medium
Frequency: Always (any compile call without explicit strict:false)
Workaround: Pass {"strict":false} explicitly when compiling via the MCP tool.
Logs/Diagnostics
-# gh-aw-metadata: {...,"agent_id":"copilot",...}+# gh-aw-metadata: {...,"strict":true,"agent_id":"copilot",...}
...
- GH_AW_COMPILED_STRICT: "false"+ GH_AW_COMPILED_STRICT: "true"
...
- - name: Enforce strict mode policy- if: ${{ vars.GH_AW_POLICY_STRICT == 'true' }}- run: |- echo "::error::GH_AW_POLICY_STRICT=true but this workflow was not compiled in strict mode. Recompile with --strict or strict: true."- exit 1
Additional Context
Other findings from this session's exploratory testing were all clean: status, logs (basic download, workflow-name filter for unknown workflow, future-date edge case), audit (successful run, invalid run_id) all returned correct results with clear, actionable error messages and no crashes.
Warning
threat detection engine error
The threat detection engine encountered an error and could not complete analysis. This is a tooling failure, not a security finding.
Details
The threat detection engine failed to produce results.
Review the workflow run logs for details.
Problem Description
The
compiletool in the agentic-workflows MCP server appears to defaultstricttotrue, even when the--strictflag is not passed. This causes the generated.lock.ymlto differ from the committed version for workflows that were originally compiled without strict mode (e.g.ace-editor.md), producing unwanted diffs and dropping theEnforce strict mode policystep.Command/Tool
printf '{"workflows":["ace-editor"]}' | agenticworkflows compile .Steps to Reproduce
git statusshows a clean working tree for.github/workflows/.agenticworkflows compilewith{"workflows":["ace-editor"]}(nostrictparam supplied).git diff .github/workflows/ace-editor.lock.yml.Expected Behavior
Without an explicit
strictparameter, compile should use the workflow's existing/previous strict setting (or defaultfalse, matchinggh aw compileCLI default), leaving the lock file unchanged when the markdown source hasn't changed.Actual Behavior
The regenerated lock file changes:
and the
gh-aw-metadataheader gains a"strict":truefield. It also removes theEnforce strict mode policystep for the previous (non-strict) mode. Simply running compile via the MCP tool without special flags mutates lock files that would otherwise be byte-identical, and could mask/alter workflow policy enforcement behavior.Environment
Impact
strict:false){"strict":false}explicitly when compiling via the MCP tool.Logs/Diagnostics
Additional Context
Other findings from this session's exploratory testing were all clean:
status,logs(basic download, workflow-name filter for unknown workflow, future-date edge case),audit(successful run, invalid run_id) all returned correct results with clear, actionable error messages and no crashes.