The Agent Skills Directory

[Command Execution]: The skill utilizes command-line tools including git and gh to manage branches, view pull request status, and interact with GitHub Actions. These operations are standard for the skill's purpose of managing CI workflows.
[Data Ingestion and Analysis]: Several steps involve downloading job logs and artifacts (e.g., using gh api or gh run download) and then parsing them with utilities like grep, sed, and tail. This process facilitates the identification of error patterns and failure causes.
[Indirect Prompt Injection Surface]: The skill defines a workflow where the agent ingests external data from CI logs and subsequently has the capability to perform write operations like git push. While the content of CI logs could potentially be influenced by external inputs (such as code from a pull request), the skill targets trusted project resources and follows established debugging procedures.
Ingestion points: Job logs retrieved via gh api and gh run view, and artifacts downloaded via gh run download (e.g., SKILL.md).
Boundary markers: Not explicitly defined in the provided commands.
Capability inventory: File system writes (redirecting logs to files), and version control operations including git add, git commit, and git push (e.g., SKILL.md).
Sanitization: The skill relies on standard shell utilities for filtering, but does not implement specific sanitization for LLM context injection.

fix-ci-failures