The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from external competitor websites, which is an untrusted source that could contain malicious instructions.
Ingestion points: External pricing pages specified by the user during research.
Boundary markers: The agent is instructed to 'Cite sources', 'preserve attribution', and 'Separate factual observations from analysis', which helps differentiate between external data and the agent's own reasoning.
Capability inventory: The skill uses sandbase_run to perform extraction and data analysis.
Sanitization: No specific sanitization or content filtering is mentioned in the prompt instructions for the extracted data.
[DYNAMIC_EXECUTION]: The workflow relies on sandbase_discover and sandbase_inspect to identify and define tools at runtime before execution via sandbase_run. This dynamic approach means the specific logic executed depends on the discovery results returned during the session.

pricing-page-analyzer