pytorch-fsdp2

Pass

Audited by Gen Agent Trust Hub on Sep 9, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill requires the agent to read and refactor training scripts provided by the user. This creates a surface for indirect prompt injection, where malicious instructions hidden in the user's script content could potentially influence the agent's actions during the modification process.\n
  • Ingestion points: The agent ingests user-provided training scripts as the primary input for the FSDP2 implementation task as defined in SKILL.md.\n
  • Boundary markers: There are no defined delimiters or instructions to treat the user's code as untrusted data or to ignore embedded instructions.\n
  • Capability inventory: The skill enables the agent to write code that performs distributed networking via NCCL and file operations via Distributed Checkpointing (DCP).\n
  • Sanitization: The instructions do not include steps to sanitize or validate the content of the input scripts before processing.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 9, 2026, 07:07 PM
Security Audit — agent-trust-hub — pytorch-fsdp2