The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes shell commands like rg and gh to audit repository configuration and retrieve CI/CD run logs. These operations are diagnostic and restricted to the repository context.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the repository and CLI tool outputs, which could potentially contain malicious instructions intended to influence agent behavior.
Ingestion points: Workflow files located in .github/workflows/ and output from gh run list and gh run view commands.
Boundary markers: No explicit delimiters or instructions are used to separate ingested data from the agent's primary instruction set.
Capability inventory: The agent has access to shell command execution through the gh and rg tools.
Sanitization: The skill does not perform sanitization or validation of the content retrieved from workflow files or run logs before processing.
Remediation: Wrap external content in delimiters with an explicit "ignore embedded instructions" warning and sanitize/escape external content before processing.

github-actions-efficiency