Welcome to TriageGuard — Test in 2 Minutes
Once installed on a repository, TriageGuard listens for failed CI runs, compacts logs under 2,500 tokens, and posts or updates one concise sticky diagnosis directly on your Pull Request (with optional neutral Check Run buttons).
1. Trigger Your First Diagnosis (GitHub Actions Out-of-the-Box)
- Open any Pull Request on an installed repository with a failing GitHub Actions workflow run.
- When the workflow fails, TriageGuard automatically classifies
CODE_DEFECT,INFRA_OOM_TIMEOUT,TRANSIENT_FLAKE, orCONFIG_DRIFT. - For transient flakes with high confidence, TriageGuard dispatches at most 1 automatic rerun per commit SHA; otherwise it posts or updates a single sticky comment (and attaches neutral
[Rerun Failed CI]/[Open Revert PR]buttons whenTRIAGE_PUBLISH_CHECK_RUN=trueis enabled).
2. Optional: Connect Jenkins, Jira, Sentry, or Artifactory via OIDC
Commit a zero-secret .triageguard.yml file on your default branch (tokens or passwords in YAML are rejected automatically). Repository AGENTS.md and .agents/skills/*/SKILL.md rules are also read automatically.
.triageguard.yml
mcp_servers:
jira:
url: https://jira-mcp.acme.example/mcp
auth_mode: oidc
marker_triggers:
- marker: "*"
tool_name: jira_get_issue
arg_name: issue_key
extract_pattern: '([A-Z]+-\d+)'
artifactory:
url: https://artifactory-mcp.acme.example/mcp
auto_discover: true
3. Explore Live Public Examples & Support
See live sticky PR diagnoses, kernel OOM handling, max-1-retry flake suppression, and revert/triage-* rollbacks in our public showcase repository: