The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external analytics APIs which serves as a potential surface for indirect prompt injection.
Ingestion points: Analytics data is fetched from https://analytics.intentsolutions.io via curl commands in the collection workflow (SKILL.md).
Boundary markers: The skill requires an 'Operating Contract' and specific 'Evidence Labels' (Observation, Comparison, Hypothesis, Recommendation) to separate raw data from agent interpretations (references/operating-contract.md).
Capability inventory: The skill utilizes Bash (curl, python3), Read, and a multi-agent routing system to process data.
Sanitization: A 'Verification Gate' is implemented to independently recalculate deltas and verify anomaly claims against site baselines before report delivery.
[DYNAMIC_EXECUTION]: The skill uses inline script execution to process structured data from the API.
Evidence: The authentication workflow uses python3 -c to extract the bearer token from the JSON login response (SKILL.md, Step 3).
Context: The Python execution is limited to a static, non-malicious one-liner for data parsing and does not process external code at runtime.

web-analytics