The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill instructions direct the agent to read sensitive local configuration files to identify keys for automation.
Evidence: The skill processes paths such as .env, .env.example, and GitHub secret references in .github/workflows/* to scope the required setup steps.
[COMMAND_EXECUTION]: The skill performs shell operations on generated script files to prepare them for user execution.
Evidence: The agent is instructed to run bash -n for syntax validation and chmod +x to make the resulting wizard executable.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from repository documentation and CI configurations to derive its setup logic, which represents a potential injection surface.
Evidence Chain:
Ingestion points: The agent reads README.md and .github/workflows/* files from the project repository to identify necessary setup stages.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are specified for these inputs.
Capability inventory: The skill is capable of writing to .env files, setting GitHub secrets via the gh CLI, and executing shell commands on local files.
Sanitization: No specific sanitization or validation logic is defined for the content extracted from these files.
[DYNAMIC_EXECUTION]: The skill generates a new executable bash script based on a provided template and user-defined logic.
Evidence: The "Author the wizard" step involves copying scripts/template.sh and populating it with dynamically generated stages before setting it as executable.

wizard