The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external sources, including search results and web scraping, through the sandbase_run tool. This creates a potential surface for indirect prompt injection attacks where malicious instructions could be embedded in the retrieved content. * Ingestion points: External content enters the agent context via the SandBase API tools as described in SKILL.md and references/sandbase-api-map.md. * Boundary markers: The skill does not explicitly define delimiter-based boundary markers for external data, though it emphasizes keeping core work local. * Capability inventory: The skill instructions do not include direct file system access or shell execution, although the agent's environment likely possesses such capabilities. * Sanitization: The workflow includes quality gates requiring independent verification of calculations and the maintenance of an audit trail to help identify and mitigate issues arising from untrusted data.

close-management