data-inconsistency-debug

Pass

Audited by Gen Agent Trust Hub on Aug 24, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is subject to Indirect Prompt Injection risks. It retrieves and analyzes structured data (JSON report definitions and data results) from an external CLI tool without implementing protective measures to ensure the agent ignores any instructions potentially embedded within that data.
  • Ingestion points: External data enters the agent's context through the outputs of ae-cli analysis report get and ae-cli analysis report-data run commands referenced in Stage One and Stage Three of the SKILL.md file.
  • Boundary markers: The instructions do not define delimiters (such as XML tags or unique string markers) to separate tool-generated content from the agent's internal reasoning or user instructions.
  • Capability inventory: The skill allows for the execution of several ae-cli sub-commands which fetch sensitive metadata and report configurations.
  • Sanitization: There are no instructions provided for the agent to validate, escape, or filter the content returned by the external tool before processing it for decision-making.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 24, 2026, 10:17 AM
Security Audit — agent-trust-hub — data-inconsistency-debug