The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection as it is designed to ingest and process untrusted data from external sources.
Ingestion points: Data enters the agent context via the WebFetch tool (fetching external URLs for audits) and local research files such as research/icp-research.md and research/product-context.md.
Boundary markers: The orchestrator uses structured Markdown headers and a JSON-like pre-writing object to organize context before dispatching to sub-agents, providing some structural separation.
Capability inventory: The skill has access to several powerful tools including Bash, WebSearch, WebFetch, Read, Grep, and Glob.
Sanitization: There are no explicit instructions to the agent to sanitize external content or ignore potential embedded instructions within fetched web data.

seo