explore
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill's documentation provides installation commands using
git cloneandnpxthat target the vendor's official GitHub repository (oil-oil/codex-explore-skill). These are standard installation methods for the vendor's resources. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and summarize large volumes of data from external, potentially untrusted codebases, creating a vulnerability surface for indirect prompt injection.
- Ingestion points: The skill reads arbitrary files and directory structures from the target codebase via explorer subagents or the
rg(ripgrep) utility. - Boundary markers: The instructions attempt to mitigate risks by requiring subagents to return only concise, structured reports in a specific table format.
- Capability inventory: The skill utilizes file system search (
rg), file reading, and the ability to spawn and instruct subagents. - Sanitization: The workflow enforces a strict "read-only" reconnaissance pass, which isolates the initial data processing from the main agent's implementation and decision-making context.
Audit Metadata