The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill is explicitly designed to handle and persist secrets (API keys, tokens) to sensitive files like .env, .env.local, and .env.production. It also instructs the agent to read these files and GitHub workflow secrets to determine required configuration, which exposes existing credentials to the agent's context.
[INDIRECT_PROMPT_INJECTION]: The skill determines its logic and execution stages by reading repository files such as .env.example, README, and .github/workflows/*. Maliciously crafted files in a repository could influence the agent's scoping phase, potentially leading to the generation of a wizard that performs unintended actions or captures data incorrectly.
Ingestion points: Processes project files (.env, .env.example, docker-compose.yml, GitHub workflows) in SKILL.md (Step 1).
Boundary markers: None provided for the reading of external configuration files.
Capability inventory: Writing bash scripts, changing file permissions (chmod +x), and using the GitHub CLI (gh secret/variable set).
Sanitization: No explicit sanitization mentioned for the manual inputs captured by the generated wizard before being written to files.
[COMMAND_EXECUTION]: The skill involves the creation of executable shell scripts and utilizes the GitHub CLI tool for administrative tasks. It specifically instructs the agent to perform chmod +x on generated files to enable execution.
[DYNAMIC_EXECUTION]: The skill follows a pattern of generating bash scripts from a local template (template.sh) and then validating or preparing them for execution. While this is the intended purpose of the tool, generating and marking scripts as executable based on project analysis is a dynamic code generation pattern.

wizard