[integrity] DIFC Integrity-Filtered Events Report — 2026-03-29 #23479
Closed
Replies: 1 comment
-
|
This discussion has been marked as outdated by Daily DIFC Integrity-Filtered Events Analyzer. A newer discussion is available at Discussion #23580. |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
-
Executive Summary
In the last 7 days, 221 DIFC integrity-filtered events were detected across 22 workflow runs spanning 10 distinct workflows. The most frequently filtered tool was
search_issues(111 events), closely followed bylist_issues(100 events). The dominant filter reason wasintegrity(152 events, 68.8%), indicating that the most common cause is content from users with no existing trust relationship (none:alltag) rather than explicitlyunapprovedcontent.The event distribution shows an increasing trend over the 3-day window — from 51 on Mar 27 to 92 on Mar 28, then moderating to 78 on Mar 29. The Auto-Triage Issues workflow accounts for 60% (132 events) of all filtered events, reflecting its broad intake of community-submitted issues. A long tail of persistent unresolved issues from ~26 unique external contributors keeps re-surfacing across multiple workflow runs, amplifying the total count.
Key Metrics
integrity(68.8%)szabta89(38 events)📈 Events Over Time
The 3-day period shows a spike on 2026-03-28 (92 events) driven by the Sub-Issue Closer and Auto-Triage workflows, which process batch queues of open issues. The dip on 2026-03-27 (51 events) likely reflects fewer workflow triggers over the weekend. Events on 2026-03-29 remain elevated at 78, suggesting a stable backlog of unresolved low-integrity issues continuing to be encountered by multiple workflows.
🔧 Top Filtered Tools
search_issues(111) andlist_issues(100) dominate, comprising 95% of all events. Both tools are used by the issue-processing workflows to enumerate open issues in the repository — the filter fires when results include issues authored by users with insufficient trust level.issue_read(7) andpull_request_read(3) represent direct lookups of specific resources that also happen to have low integrity. All filtered events originate from a single MCP server:github.🏷️ Filter Reasons and Tags
All 221 events carry the
none:allintegrity tag, indicating issues from users with no established trust association. 69 events (31.2%) additionally carryunapproved:all, flagging content that was explicitly reviewed and not approved. There are no secrecy-tagged events — this is a pure integrity signal (untrusted contributor content), not a confidentiality concern. No false-positive pattern (e.g., internal team members being filtered) is visible in the data.📋 Per-Workflow Breakdown
📋 Per-Server Breakdown
All DIFC filtering originates exclusively from the
githubMCP server — no other servers are involved.👤 Per-User Breakdown (Top 26)
🔍 Per-User Analysis
The top 4 users (
szabta89,danielmeppiel,deyaaeldeen,mnkiefer) account for 54% of all filtered events (119/221). These are repeat offenders whose open issues appear across multiple workflow runs and multiple tool calls within each run. No bot accounts or automated actors are among the blocked users — all are human contributors whose issues either lack theapprovedintegrity label or are authored by users withNONEorCONTRIBUTORGitHub association. TheEsomoire-consultancy-Companyaccount (5 events) is notable as a NONE-association account that persistently appears in search results forDevandDaily Documentation Updaterworkflows.💡 Tuning Recommendations
Approve or close long-standing community issues: Issues from
szabta89,danielmeppiel,deyaaeldeen, andmnkieferare blocked across 22+ runs. Review these open issues and either apply theapprovedintegrity label (if valid) or close stale ones — this would eliminate ~54% of filtered events immediately.Add
none:allpass-through for read-only triage workflows: The Auto-Triage Issues and Sub-Issue Closer workflows need to read community issues to label them — by design. Consider configuring these workflows with a lower minimum-integrity threshold (e.g.,minimum_integrity: none) so they can process issues from external contributors without being blocked. This is safe since both workflows write labels, not code.Investigate
Devworkflow: TheDevworkflow (13 events, 2 runs, bothfailure) does not need to triage community issues. Review its prompt to ensure it is not inadvertently fetching untrusted issue content when reporting on repository state.Whitelist or fast-track
CONTRIBUTORassociation users: 69 events (31%) carryunapproved:alland belong toCONTRIBUTOR-associated accounts (szabta89,mnkiefer,eaftan). These contributors have merged code before — their issues should likely be reviewed promptly forapprovedlabeling.Monitor Smoke workflows:
Smoke CopilotandSmoke Claudeeach had filtered events (2 and 1 respectively), which is unexpected for smoke/validation workflows. Review whether the smoke workflow prompts include issue reads that could be tightened.Filter stability check for Mar 28 spike: The 92-event spike on 2026-03-28 coincides with the Sub-Issue Closer (19 events) and Auto-Triage Issues (multiple runs). No new issue volume is evident — the spike appears driven by the same backlog of low-integrity issues being iterated on multiple workflow schedules. No further investigation needed unless the spike persists beyond this week.
Generated by the Daily Integrity Analysis workflow
Analysis window: 2026-03-27 to 2026-03-29 | Repository: github/gh-aw
Run: https://github.com/github/gh-aw/actions/runs/23717775390
Beta Was this translation helpful? Give feedback.
All reactions