The Agent Skills Directory

[SAFE]: The skill implements a robust 'Guardrails' section that explicitly prohibits the use of secrets, credentials, or PII. It also forbids providing exploit guidance, limiting vulnerability discussions to defensive purposes only.
[SAFE]: The 'Evidence Protocol' establishes high standards for data integrity, requiring multi-source triangulation, freshness checks (≤90 days), and confidence ratings for all claims.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to fetch and process untrusted data from external blockchain documentation and GitHub repositories. It mitigates the risk of indirect prompt injection through its structured evaluation framework and requirements for primary and independent verification. Capability inventory: Uses WebSearch and WebFetch for data ingestion; no write capabilities or execution of external data detected.

research-blockchain