The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands using the gh (GitHub CLI) and git binaries to perform repository operations such as cloning, listing, and creating pull requests.
[EXTERNAL_DOWNLOADS]: The documentation provides links and instructions to download the official GitHub CLI from trusted sources including cli.github.com and standard package managers like Homebrew, Winget, and Scoop.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection as it ingests data from external sources (GitHub repository lists, issue titles, and PR descriptions) and interpolates the user's question directly into the subagent prompt.
Ingestion points: External data is ingested via gh repo list, gh pr list, and gh issue list in SKILL.md.
Boundary markers: The agent is instructed to show the exact command before running it and wait for user confirmation ('yes/no'), which serves as a human-in-the-loop boundary.
Capability inventory: The skill can execute shell commands (gh, git), access the network (via gh), and write to a local log file (~/.ai-software-agency/audit.log).
Sanitization: The agent is instructed to interpret raw error messages and present them to the user, providing transparency into command execution.
[DATA_EXPOSURE]: The skill includes instructions to log operations to a local file (~/.ai-software-agency/audit.log), which is a standard auditing practice and does not involve exfiltration to external domains.

specialist-engineering-github-operator