jolt-physics

Pass

Audited by Gen Agent Trust Hub on Nov 6, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill implements significant security safeguards, particularly within the scripts/preflight.py helper. This script uses secure POSIX file operations (O_NOFOLLOW, dir_fd) to prevent symlink-based directory traversal attacks. It also explicitly blocks access to sensitive file paths and stems (e.g., credentials, secrets, tokens, and cloud configuration directories like .aws or .azure).
  • [COMMAND_EXECUTION]: The instructions direct the agent to run a local verification script, scripts/preflight.py, to perform offline integrity checks on repository files. This is a controlled execution of a vetted local script intended to enhance security by ensuring source file provenance.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to process and integrate external source code, creating a potential surface for indirect prompt injection from untrusted repositories.
  • Ingestion points: Local files within a user-provided source directory (e.g., README.md, LICENSE, and various C++ source files).
  • Boundary markers: The skill requires a 'preflight' check that validates the bytes of specific files against a list of pinned SHA256 hashes in references/source.json before performing integration tasks.
  • Capability inventory: The skill is granted Read, Write, Edit, and Bash permissions to facilitate code integration and build planning.
  • Sanitization: The use of the preflight.py script ensures that the most critical files being analyzed match a known-good upstream state, effectively mitigating the risk of processing malicious instructions embedded in those specific files.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 6, 2026, 08:52 AM