-
Notifications
You must be signed in to change notification settings - Fork 25
Actions Manager
GitHub Actions command center -- view workflow runs, read logs, re-run failed jobs, manage workflows, and debug CI failures entirely from the editor. Bypasses the deeply nested, visually-dependent Actions UI that is largely inaccessible to screen readers.
The GitHub Actions web UI is one of the most difficult parts of GitHub to navigate with a screen reader. Workflow runs are nested inside expandable sections, log output is rendered in a virtual-scroll container that announces nothing useful, and re-running a failed job requires finding small icon buttons with no accessible labels.
The actions-manager makes every Actions operation available through plain-text commands. You can list runs, read logs, re-run failures, and debug CI issues without ever opening the browser. All output is structured as tables and annotated text that works with any assistive technology.
Use actions-manager when:
- A CI run has failed and you need to read the failure logs
- You want to re-run a specific failed job without re-running the entire workflow
- You need to check which workflows ran on a branch or pull request
- You want to download or inspect build artifacts
- You need to enable, disable, or trigger a workflow dispatch
- You want a summary of recent CI health across your repository
@actions-manager show recent workflow runs
@actions-manager why did CI fail on this PR?
@actions-manager re-run the failed jobs on run 12345
@actions-manager show logs for the "test" job
@actions-manager list artifacts from the latest build
/actions-manager show failed runs
/actions-manager read logs for run 12345
/actions-manager re-run failed jobs
/actions-manager list workflow files@github-hub why did CI fail?
@github-hub re-run the failed build
Expand full capability reference
| Category | What it does | API |
|---|---|---|
| List runs | Show recent workflow runs filtered by branch, event, status, or actor | REST API workflow runs |
| Run details | Display timing, status, trigger event, and commit info for a run | REST API run details |
| Job details | List jobs within a run with individual status and duration | REST API workflow jobs |
| Log reading | Download and display job logs with failure annotation extraction | REST API job logs |
| Re-run failed | Re-run only the failed jobs in a run, not the entire workflow | REST API re-run endpoint |
| Re-run all | Re-run all jobs in a workflow run | REST API re-run endpoint |
| Cancel run | Cancel an in-progress workflow run | REST API cancel endpoint |
| Artifacts | List, download, and inspect artifacts from a workflow run | REST API artifacts |
| Workflow management | Enable, disable, list, and trigger workflows via dispatch | REST API workflow endpoints |
| Failure diagnosis | Extract error messages, failed steps, and annotations from logs | Log parsing and annotation API |
-
Run identification:
Fetching recent failed runs for main branch... Found 1 failed run: #4521 "CI" triggered by push (3 minutes ago) -
Job breakdown. The agent lists all jobs in the run with their status. Passed jobs are noted briefly; failed jobs are highlighted with the failing step name.
-
Log extraction. The agent downloads the failed job's logs and extracts the relevant error. You see the actual failure message, not pages of setup output.
-
Diagnosis. Based on the error, the agent suggests what went wrong (test failure, dependency issue, linting error, timeout) and what to do next.
-
Re-run option. If the failure looks transient (network timeout, flaky test), the agent offers to re-run only the failed jobs.
- The agent identifies the most recent failed run (or you specify one)
- It confirms which jobs failed and which passed
- Only failed jobs are re-run, saving time and compute
- The agent reports the new run URL so you can track progress
- The agent queries the specified run for artifacts
- Each artifact is listed with its name, size, and expiration date
- You can request a download of any artifact
- The agent extracts and summarizes artifact contents when possible
Expand agent connections
| Handoff | When | Agent |
|---|---|---|
| Route from hub | "check CI status" or "why did the build fail" | github-hub |
| PR context | After diagnosing a CI failure on a PR, hand off for code review | pr-review |
| Issue creation | When a CI failure reveals a bug, create an issue to track it | issue-tracker |
| Agent | Relationship |
|---|---|
| github-hub | Parent router -- delegates CI and workflow commands here |
| pr-review | Reviews PRs whose CI status this agent inspects |
| issue-tracker | Creates issues for bugs discovered through CI failures |
| repo-manager | Scaffolds the workflow files that this agent monitors |
| security-dashboard | Triages security alerts that may surface through CI scans |
Show troubleshooting help
"No workflow runs found." Runs are scoped to a repository. Make sure you are in a git repository and the remote points to GitHub. If filtering by branch, check the branch name is correct.
"Cannot re-run -- the run is too old." GitHub only allows re-runs within 30 days of the original run. For older failures, trigger a new run instead: "dispatch the CI workflow on main."
"Logs are too large to display." For very large log files, the agent extracts only the failed step output. Ask: "show only the error lines from run 12345" for a focused view.
"Permission denied when re-running."
You need write access to the repository to re-run workflows. Check your permissions with gh auth status.
- Accessibility Lead
- Web Accessibility Wizard
- Document Accessibility Wizard
- Alt Text and Headings
- ARIA Specialist
- Contrast Master
- Forms Specialist
- Keyboard Navigator
- Link Checker
- Live Region Controller
- Modal Specialist
- Tables Data Specialist
- Word Accessibility
- Excel Accessibility
- PowerPoint Accessibility
- PDF Accessibility
- Office Scan Config
- PDF Scan Config
- Testing Coach
- WCAG Guide