The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest and analyze external data from CI logs and GitHub PR metadata, creating a vulnerability surface where malicious content in those sources could influence agent behavior.
Ingestion points: PR metadata (title, labels, author) via gh pr view, PR diffs via gh pr diff, and log artifacts downloaded from GitHub Actions via gh run download.
Boundary markers: The instructions do not define clear delimiters or provide instructions for the agent to ignore embedded commands within the processed log files.
Capability inventory: The skill uses gh CLI for repository interaction, local shell tools (grep, sed, wc) for log processing, and a Python script for remote repository interaction.
Sanitization: There are no documented steps for sanitizing or escaping the content of the logs before processing.
[COMMAND_EXECUTION]: The skill provides instructions for triggering an internal CI pipeline using tools/trigger_internal_ci.py, which performs a force-push to a remote GitLab repository. While the skill includes a prominent warning about the destructive nature of this command and mandates a --dry-run preflight check, it remains a high-risk operation that requires careful oversight.

mcore-cicd