The Agent Skills Directory

[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it generates executable code from untrusted user input and repository data.
Ingestion points: Manual procedure descriptions and existing repository files like .env.example or CI workflows.
Boundary markers: The skill requires user confirmation of the stage list and captured values before generating the script.
Capability inventory: The generated script has the ability to write to the file system (.env) and interact with the GitHub CLI (gh) to set secrets.
Sanitization: Relies on manual human review and static analysis tools like shellcheck or bash -n.
[COMMAND_EXECUTION]: The skill generates scripts that execute system commands to interact with the environment and external services.
Evidence: The template in scripts/template.sh uses commands like gh secret set, gh variable set, tput, and various platform-specific browser openers.
[DATA_EXFILTRATION]: The generated scripts are designed to collect and transmit sensitive information as part of their core functionality.
Evidence: Functions like ask_secret and set_secret are used to capture user credentials and upload them to GitHub Actions repositories.
[EXTERNAL_DOWNLOADS]: The skill is derived from and references an external GitHub repository for its logic and template.
Evidence: References to github.com/mattpocock/skills are included in README.md and SKILL.md.

wizard