contract-red-flags
Pass
Audited by Gen Agent Trust Hub on Aug 13, 2026
Risk Level: SAFEPROMPT_INJECTIONNO_CODE
Full Analysis
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests untrusted text (contracts) provided by the user without establishing security boundaries.
- Ingestion points: Untrusted data enters the agent context via the user-provided contract text described in the 'Required Inputs' section of SKILL.md.
- Boundary markers: Absent. The instructions do not define delimiters for input text or specify that the agent should ignore instructions embedded within the contract.
- Capability inventory: No capability for code execution, file modification, or network access was identified as the skill does not contain scripts or tools.
- Sanitization: Absent. There is no mention of input validation or content filtering to mitigate the risk of malicious instructions in the processed data.
Audit Metadata