reading-comprehension-strategy-selector

Pass

Audited by Gen Agent Trust Hub on Apr 2, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection where malicious instructions embedded in input data could influence the agent's behavior.
  • Ingestion points: Untrusted data enters the agent context via the {{text_description}}, {{reader_challenge}}, {{student_profiles}}, and {{text_extract}} variables in the SKILL.md prompt template.
  • Boundary markers: The prompt lacks robust delimiters (such as XML tags or unique string markers) to separate the instruction set from the user-provided text content, relying only on simple bold labels.
  • Capability inventory: The skill is configured with disable-model-invocation: false and does not restrict tool access in the frontmatter, meaning successful injection could potentially leverage any tools available in the agent's environment.
  • Sanitization: The instructions do not include validation steps or directives to the model to ignore instructions found within the processed text extracts.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 2, 2026, 10:53 PM