The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's workflow involves scanning local repository files to determine the necessary setup variables and stages.
Ingestion points: The agent is instructed in SKILL.md to read files such as .env, .env.example, README, docker-compose*, and .github/workflows/* to identify values the wizard must produce.
Boundary markers: The instructions lack explicit boundary markers or directives to the agent to ignore potentially malicious instructions embedded within the processed repository data.
Capability inventory: The resulting script possesses capabilities to write to the local file system (write_env), modify GitHub repository secrets and variables via the gh CLI (set_secret, set_var), and execute system-level browser opening commands (open_url).
Sanitization: The skill does not prescribe sanitization or validation for data extracted from the repository files before incorporating it into the generated bash script.
[COMMAND_EXECUTION]: The core function of the skill is the generation and execution of shell scripts.
The instructions direct the agent to author a script based on template.sh, verify it with bash -n, and make it executable using chmod +x.
The provided template.sh library functions rely on executing shell commands including grep, mv, touch, tput, and external binaries like the GitHub CLI (gh). While these are necessary for the skill's intended purpose, they represent a broad execution surface.

wizard