The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes user-supplied brand information stored in local Markdown files across multiple sessions to maintain interview continuity.
Ingestion points: Reads module files (e.g., modules/10_purpose-why.md) and state.json from the local brand-identity directory in the SKILL.md start protocol.
Boundary markers: The module files use Markdown headers (## Raw) to separate user input, but there are no explicit instructions for the agent to ignore potentially malicious commands that might be embedded in the user's past verbatim responses.
Capability inventory: The skill performs local file writes to state.json, modules/*.md, and founders/*.md to record interview progress.
Sanitization: The skill implements strong validation for participant names (alphanumeric and hyphens only), module filenames (fixed sequence 10-90), and output paths (absolute paths only, rejecting .. segments) in the Multi-founder mode section of SKILL.md.

brand-discovery