arc-contract-review
Pass
Audited by Gen Agent Trust Hub on Sep 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONNO_CODE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted external data in the form of legal contracts provided by users.
- Ingestion points: The agent is instructed to read and analyze contract text from user prompts or provided files (e.g.,
sample-saas-agreement.md). - Boundary markers: The instructions in
SKILL.mddo not define specific delimiters (like XML tags or triple backticks) to isolate the untrusted contract text from the agent's system instructions. - Capability inventory: The skill focuses on reasoning, extraction, and text generation. It does not define custom tools or explicit shell/network commands, though it is intended for use in environments like Claude Code that possess those capabilities.
- Sanitization: There is no evidence of automated filtering or sanitization of the input text to prevent embedded instructions from influencing agent behavior.
- [NO_CODE]: The skill consists entirely of markdown documentation, instructional templates, and evaluation configurations. It does not include any executable code (e.g.,
.py,.js,.shfiles) or package dependencies.
Audit Metadata