The Agent Skills Directory

[SAFE]: The skill implements a 'Framework Advisor' role that uses structured logic (Engine 1 for routing and Engine 2 for heft) to recommend thinking processes. It includes robust defensive instructions to handle untrusted user data, specifically requiring the agent to 'mirror' input and build a 'source ledger' to ensure all claims are grounded in provided text. It explicitly directs the agent to treat potential prompt injection markers (e.g., 'ignore the above') found in user content as data to be analyzed rather than instructions to be followed, which is a key security best practice for LLM agents. References to external repositories and internal build scripts are associated with the vendor 'product-on-purpose' and do not involve remote code execution or unauthorized data access.

think-framework-advisor