reflexion-pattern
Pass
Audited by Gen Agent Trust Hub on Jul 27, 2026
Risk Level: SAFEPROMPT_INJECTIONNO_CODE
Full Analysis
- [PROMPT_INJECTION]: The
Critic prompt templateinSKILL.mddefines a structure that interpolates untrusted data, creating a surface for indirect prompt injection.\n - Ingestion points: The
{trajectory}and{task}variables are used to build the prompt for the critic agent node.\n - Boundary markers: The template lacks explicit delimiters (such as XML tags or unique markers) to isolate untrusted content from instructions.\n
- Capability inventory: The skill does not provide an implementation; it is a design pattern for an 'Actor' that would execute tasks, which could involve various capabilities depending on the user's setup.\n
- Sanitization: There are no instructions or mechanisms provided for sanitizing or validating the input data before it is processed by the LLM.\n- [NO_CODE]: The skill consists exclusively of markdown documentation and non-functional code snippets used as examples. It does not include any executable code, binary components, or automation logic that would be executed by the agent environment.
Audit Metadata