The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user input via setup questions and writes the data directly to local configuration files without sanitization or boundary delimiters, potentially affecting downstream skills that read these files.
Ingestion points: User responses to the seven shared questions (name, title, company, ICP segments, exclusions, and customer stories) provided during the setup process.
Boundary markers: Absent; user input is merged directly into JSON fields in brand-config.json and Markdown sections in SOUL.md without delimiters or "ignore instructions" wrappers.
Capability inventory: The skill possesses Read and Write capabilities for brand-config.json, SOUL.md, and the ./gtm/ directory, as well as Glob for file discovery.
Sanitization: No validation or sanitization is performed on the user's answers before they are persisted to the filesystem.

setup