mode-selection
Pass
Audited by Gen Agent Trust Hub on Mar 27, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill establishes a delegation protocol where tasks and payloads are passed between specialized modes (such as
/githubberfor Git operations or/coder-srfor complex engineering). This architecture creates an indirect prompt injection surface where instructions from untrusted data could be propagated across different agent contexts. - Ingestion points: Data enters the system through the
taskandpayloadfields described in the 'Mode selection strategy' inSKILL.md. - Boundary markers: None are specified; the instructions do not mandate the use of delimiters or 'ignore' warnings when interpolating data into payloads for other modes.
- Capability inventory: The strategy orchestrates access to modes with significant capabilities, including version control (
/githubber), complex code execution (/coder-sr), and file system operations (/tasky). - Sanitization: The protocol lacks explicit requirements for validating, escaping, or sanitizing the 'Instructions' or 'Context' passed between modes.
Audit Metadata