meta-cognition-parallel
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input from the
$ARGUMENTSvariable and interpolates it directly into prompts intended for sub-agents or internal analysis sections. - Ingestion points: User-provided Rust questions enter the agent context through the
$ARGUMENTSparameter inSKILL.md. - Boundary markers: The skill uses Markdown headers (e.g.,
## User Query) to separate user content from the analyzer's instructions, though it lacks strict enforcement or "ignore previous instructions" safety guardrails. - Capability inventory: The skill is limited to reading local agent configuration files and outputting text; it does not have capabilities for network operations, file writing, or system command execution.
- Sanitization: No explicit filtering or sanitization is performed on the user input before it is used to construct sub-agent prompts.
Audit Metadata