The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill scopes manual procedures by reading repository files such as README, .env.example, and GitHub workflow files. Maliciously crafted content in these files could potentially influence the agent to generate a script with unintended or harmful steps.
Ingestion points: Analyzes local repository configuration and documentation files in SKILL.md Step 1 to define setup stages.
Boundary markers: No specific delimiters or safety instructions are used when processing external file content to distinguish instructions from data.
Capability inventory: Generated scripts possess the ability to write to the file system (.env), manage GitHub repository secrets via the gh CLI, and execute system commands to open URLs.
Sanitization: The skill lacks explicit sanitization logic for data extracted from repository files before it is used to define script stages.
[COMMAND_EXECUTION]: The skill generates bash scripts that execute local commands for opening browsers and interacting with the GitHub CLI (gh). While these functions are part of the intended developer workflow, the generation of executable scripts based on repository context requires user verification before execution.
[DATA_EXFILTRATION]: The skill specifically targets sensitive file paths like .env, .env.example, and GitHub Secrets. While its primary function is to manage these credentials for the user, it involves the ingestion and interpolation of high-value secrets into generated scripts.

wizard