Warning 1 — Recurring detection job failures (8/30 scheduled agentic runs, 26.7%)

The detection job failed in 8 distinct workflows. In each case, downstream jobs (safe_outputs, update_cache_memory, push_repo_memory) were skipped as a result, preventing non-compliant output from being published. This behavior is consistent with the DIFC pipeline operating as designed (detection failure → block output), but the frequency across diverse workflows warrants investigation to distinguish intentional policy blocks from false positives or detection misconfigurations.

Affected runs:

In all 8 cases the job sequence was: activation ✅ → agent ✅ → detection ❌ → safe_outputs ⏭ → conclusion ✅

This means the agents completed execution and uploaded artifacts before detection ran. The detection failure correctly blocked the final write/publish step.

Warning 2 — DIFC secret-scope filtering blocked audit agent access to logs and artifacts

All attempts to download workflow artifacts or retrieve job logs were blocked by DIFC:

[DIFC] 1 item(s) in this response were removed by integrity policy and are not shown:
resource:actions_get (Resource 'resource:actions_get' has secrecy requirements that
agent doesn't meet. The agent is not authorized to access [secret]-scoped data.)

This prevented direct inspection of rpc-messages.jsonl, mcp-gateway.log, and firewall-audit-logs for DIFC event counts, filtered item ratios, guard errors, and scope violations. While the filtering itself confirms the DIFC system is correctly labeling these resources as [secret], it limits the depth of this automated audit. A privileged audit agent (or manual review) would be needed to inspect the actual integrity tags and filtered counts.

Info 1 — DIFC pipeline is active and enforcing access control

All 7 artifact download attempts and all job log requests from this audit agent were correctly filtered with [secret]-scope DIFC labels. This confirms the MCP Gateway's DIFC enforcement is intercepting GitHub API calls made by this agent and applying correct secrecy policies.

Info 2 — 30 action_required runs are pending approval (no execution, no DIFC data)

All 30 issue_comment/PR-triggered runs from the last 24 hours have action_required status — they are awaiting human approval before execution. These include workflows such as Security Review Agent, Grumpy Code Reviewer, Resource Summarizer Agent, ACE Editor Session, Mergefest, Scout, PR Nitpick Reviewer, Q, Archie, Plan Command, Content Moderation, AI Moderator, and others. No DIFC data is available for these runs as they have not yet executed.

Info 3 — 30 push-triggered CI runs are not in DIFC scope

Push event runs (License Compliance Check, Doc Build - Deploy, CI) are infrastructure/build workflows without the agentic MCP pipeline (no activation/agent/detection job structure). All 30 completed successfully. These are not expected to have DIFC events.

Info 4 — 22/30 scheduled agentic runs completed the full DIFC pipeline successfully

Successful workflows (detection passed) include:

• Daily Security Red Team Agent (×1)
• Issue Monster (×8)
• Glossary Maintainer (×1)
• Daily Team Status (×1)
• Dependabot Dependency Checker (×1)
• Contribution Check (×1)
• Go Fan (×1)
• PR Triage Agent (×1)
• CI Cleaner (×1)
• Bot Detection (×1)
• CodeQL Security Analysis (×1)
• Code Simplifier (×1)
• Agentic Maintenance (×2 — different structure, no detection job)
• Doc Build - Deploy (×1 — no detection job)