The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits a vulnerability to indirect prompt injection by design, as it captures untrusted user input and writes it to a persistent file.
Ingestion points: The skill ingests data through user responses during the interview phase (via the AskUserQuestion tool) and reads existing content from SPEC.md within the SKILL.md workflow.
Boundary markers: There are no explicit delimiters, markers, or "ignore instructions" directives defined to separate user-provided content from the skill's structural prompts in the generated SPEC.md file.
Capability inventory: The skill utilizes Read, Write, and Edit tools to modify files on the local file system based on the interview outcomes.
Sanitization: The instructions do not specify any validation, escaping, or filtering of the user's input before it is interpolated into the markdown template and saved to the disk.

spec