residues

Pass

Audited by Gen Agent Trust Hub on Apr 5, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill relies on the Bash tool to perform mathematical computations by invoking Python scripts (scripts/sympy_compute.py, scripts/z3_solve.py) with arguments constructed from user-provided mathematical expressions.
  • [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface through a command injection vulnerability where untrusted user input is interpolated into shell commands.
  • Ingestion points: User-provided function expressions (f(z)) and singularity points (z0) defined in both SKILL.md and SKILL.v6.md.
  • Boundary markers: The command templates lack boundary markers or explicit instructions to the agent to treat user input as non-executable data.
  • Capability inventory: The agent uses the Bash tool to execute uv run python commands, which can be exploited if the interpolated strings contain shell metacharacters.
  • Sanitization: No sanitization, escaping, or validation logic is specified for the user-supplied strings before they are interpolated into the shell commands.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 5, 2026, 01:57 PM