The Agent Skills Directory

[PROMPT_INJECTION] (LOW): The skill directly interpolates the user-provided idea argument into the OpenAI prompt in src/index.ts without delimiters or sanitization. A malicious input could attempt to subvert the 'investor-ready' instructions to generate unrelated or harmful content.
[INDIRECT_PROMPT_INJECTION] (LOW):
Ingestion points: User input via the idea argument in src/cli.ts.
Boundary markers: Absent. The input is directly concatenated: Create pitch deck content for: ${idea}.
Capability inventory: File system write access via fs.writeFileSync in src/cli.ts and outbound network access to the OpenAI API.
Sanitization: Absent. There is no validation or escaping of the input string before it is sent to the LLM.
[DATA_EXPOSURE] (LOW): The CLI tool allows users to specify an output file path via the -o or --output flag. If used by an automated agent without path validation, this could be used to overwrite sensitive files on the local system.

pitch-gen