explore-deep
Pass
Audited by Gen Agent Trust Hub on Jul 21, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it is designed to read and process arbitrary content from the project codebase. This is inherent to the skill's primary purpose for codebase exploration.
- Ingestion points: Untrusted data enters the agent context through the use of Glob, Grep, and Read tools during the exploration phase in SKILL.md.
- Boundary markers: The instructions do not mandate the use of delimiters or specific warnings to ignore instructions embedded within the codebase files.
- Capability inventory: The subagent possesses broad read access to the filesystem and the ability to write to the EXPLORE.md artifact.
- Sanitization: No explicit sanitization, validation, or escaping of external content is specified before the data is processed or persisted.
- [SAFE]: The skill uses dynamic shell command execution to retrieve environment information such as the current git branch and project root. These are standard development operations used appropriately for orienting the subagent within the repository context. Evidence: Shell command substitutions in the Pre-computed context section of SKILL.md.
Audit Metadata