The Agent Skills Directory

[COMMAND_EXECUTION]: The skill generates bash scripts and provides instructions to make them executable using chmod +x. This is an intended function for creating runnable utility scripts.
[INDIRECT_PROMPT_INJECTION]: The skill analyzes local repository files to determine setup stages, creating a potential surface for indirect instructions.
Ingestion points: Data is ingested from .env, .env.example, README, and GitHub workflow files in the local repository (SKILL.md).
Boundary markers: Absent; no explicit delimiters are used to isolate the ingested repository content during the scoping phase.
Capability inventory: The generated script is capable of writing to the local filesystem and managing GitHub repository secrets using the gh CLI (template.sh).
Sanitization: Absent; the process relies on the agent's static code verification and a human-in-the-loop review before the script is executed.
[DATA_EXFILTRATION]: The generated setup scripts include functionality to upload credentials to GitHub Actions secrets via the gh CLI. This interaction with GitHub's official service is the intended purpose of the skill's secret management feature.

wizard