best-practice-research
Pass
Audited by Gen Agent Trust Hub on Jul 29, 2026
Risk Level: SAFE
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill uses
WebSearchandWebFetchtools to gather evidence from official documentation, release notes, and standards on the internet. - [COMMAND_EXECUTION]: Repository inspection tools such as
Read,Grep,Glob, andBashare used to gather local context (versions, configuration, and integration points) to ground the research findings. - [PROMPT_INJECTION]: The skill ingests untrusted data from external websites via
WebSearchandWebFetch(ingestion point inSKILL.md), which creates a surface for indirect prompt injection as there are no explicit boundary markers or sanitization mentioned. However, the risk is mitigated by the inventory of capabilities being limited to read-only repo inspection tools (Bash,Read,Grep,GlobinSKILL.md) and explicit instructions that the skill must remain terminal and never modify repository state.
Audit Metadata