parallel-agent-contracts
Warn
Audited by Gen Agent Trust Hub on Feb 15, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION] (MEDIUM): The skill instructs agents to run shell commands (
grep,npx tsc) to verify types and avoid duplication. In an implementation environment, these commands are powerful and could be abused if the agent is manipulated. - [EXTERNAL_DOWNLOADS] (LOW): The use of
npx tsccan trigger downloads from the npm registry if the TypeScript package is not already present in the environment. Per [TRUST-SCOPE-RULE], npm is a trusted registry, but the dynamic nature ofnpxis worth noting. - [INDIRECT_PROMPT_INJECTION] (MEDIUM):
- Ingestion points: The
Prompt Templateaccepts[Description]and[Actual task description]as input. - Boundary markers: No boundary markers or 'ignore embedded instructions' warnings are present to isolate the task description from the agent instructions.
- Capability inventory: The agent is explicitly granted the capability to execute shell commands (
bashblocks). - Sanitization: No sanitization is performed on the task description or type names before they are used in shell commands, potentially allowing a malicious task description to append commands (e.g., via shell metacharacters in a type name).
Audit Metadata