safari-explore
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill directs the agent to execute various shell commands for environment discovery and auditing. This includes using 'ls' for directory listing, 'gf' for code searching and usage analysis, and 'uv run' to invoke a project-local tool named 'glimpse' for capturing rendered page states.
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of arbitrary source code and web content for review, creating a surface for indirect prompt injection where malicious instructions within the audited files could influence agent behavior.
- Ingestion points: Source code files read during the 'Observe' phase and local web pages captured via the 'glimpse' tool.
- Boundary markers: The instructions do not define specific delimiters or guardrails to differentiate between audited data and agent instructions.
- Capability inventory: The skill has access to file system discovery ('ls'), code search ('gf'), tool execution ('uv run'), and file creation/writing ('docs/safaris/').
- Sanitization: There is no explicit requirement for the agent to sanitize or escape the content of the files it reads before processing them into findings or design specs.
Audit Metadata