The Agent Skills Directory

Command Execution: The skill utilizes the GitHub CLI (gh) and Git to manage branches, inspect pull request status, and trigger workflow reruns. These tools are used within their standard context to facilitate local development and troubleshooting.- External Data Ingestion: The skill retrieves and processes external content from GitHub Actions logs and job artifacts. This data is essential for identifying error patterns but originates from external CI processes.- Indirect Prompt Injection Surface: The skill is designed to parse and interpret text from CI logs, which constitutes an ingestion of untrusted data into the agent's context.
Ingestion points: Logs are retrieved via the GitHub API and the gh run view command, often stored in temporary files like $TMPDIR/ci-job-log.txt.
Boundary markers: No specific delimiters or safety instructions are used to separate the log content from the agent's core operating instructions.
Capability inventory: The skill includes commands that allow the agent to perform write operations, such as git commit and git push, as well as administrative actions like gh run rerun.
Sanitization: The skill uses filtering tools like grep and sed to focus on specific log segments, though the raw input is processed without prior validation.- Targeted Repository Reference: Command examples in the skill include a reference to the microsoft/vscode repository. While appropriate for the author's context, this specific path is hardcoded in example API calls.

fix-ci-failures