The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input without sufficient isolation. 1. Ingestion points: The startup idea string is taken directly from command-line arguments in src/cli.ts. 2. Boundary markers: The input is interpolated directly into a prompt string in src/index.ts without delimiters or specific instructions to the LLM to ignore potentially malicious embedded commands. 3. Capability inventory: The skill utilizes fs.writeFileSync in src/cli.ts to save LLM output to the local file system. 4. Sanitization: There is no evidence of input validation, filtering, or escaping for the user-provided startup idea string before it is processed by the language model.
[COMMAND_EXECUTION]: The documentation instructs users to execute shell commands for installation and to set environment variables containing sensitive API keys.

pitch-gen