The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from case notes, emails, and dashboards to create structured proof assets, creating a surface for embedded instructions.
Ingestion points: Data is read from gtm/proof.json, gtm/offer-stack.json, and direct user inputs like customer testimonials.
Boundary markers: The instructions do not specify the use of delimiters or ignore-instructions to isolate untrusted data from the agent's logic.
Capability inventory: Access is limited to the Read and Write tools within the ./gtm/ directory and a specific Python script via Bash.
Sanitization: The scripts/score_proof.py script validates asset structure and content logic but does not screen for prompt injection patterns.

proof-assets