engineering-context-init
Pass
Audited by Gen Agent Trust Hub on Oct 3, 2026
Risk Level: SAFEDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DYNAMIC_EXECUTION]: The skill instructions in
SKILL.mdadvise the agent to useexec(open('init-script.py').read())for environment setup, which involves dynamic execution of the provided script file. - [INDIRECT_PROMPT_INJECTION]: The skill provides an attack surface for indirect prompt injection via data processing in helper functions.
- Ingestion points: Untrusted user data can be passed as arguments to the
atm_properties,reynolds_number, andmach_numberfunctions ininit-script.py. - Boundary markers: No boundary markers or ignore-instruction warnings are used to delimit external data from the prompt logic.
- Capability inventory: Capabilities are restricted to scientific calculations and library configuration in
init-script.py; no file-writing, network, or subprocess calls are present. - Sanitization: Input values are processed directly in mathematical formulas without explicit sanitization or validation.
Audit Metadata