meta-cognition-parallel
Pass
Audited by Gen Agent Trust Hub on Oct 10, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user input via the
$ARGUMENTSvariable and interpolates it directly into the execution context for parallel sub-agents. - Ingestion points: The user's Rust question and code snippets are captured from
$ARGUMENTSinSKILL.md. - Boundary markers: The skill uses simple Markdown headers (
## User Query) to delimit input, which provides minimal protection against adversarial input designed to override the sub-agent's instructions. - Capability inventory: The skill triggers sub-agents using the
Taskabstraction to perform cognitive analysis and synthesis. - Sanitization: There is no evidence of input validation, escaping, or filtering of the user-provided arguments before they are passed to the sub-agents.
- [DYNAMIC_EXECUTION]: The skill dynamically generates instruction sets for sub-agents at runtime.
- Evidence: In 'Agent Mode', the skill constructs prompts by reading and concatenating local Markdown files (e.g.,
../../agents/layer1-analyzer.md) with raw user input. - The execution flow relies on the agent's ability to read external files and initiate new tasks based on those contents.
Audit Metadata