moonbit-c-binding

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill includes instructions that direct the agent to inject a directive into its output plan files, mandating that subsequent agents load this skill to ensure persistent influence over multi-agent workflows.\n
  • Ingestion points: SKILL.md (Plan Mode Instructions)\n
  • Boundary markers: Absent in the injected text.\n
  • Capability inventory: subprocess.run calls in scripts/run-asan.py.\n
  • Sanitization: Absent.\n- [COMMAND_EXECUTION]: The provided validation script scripts/run-asan.py executes system shell commands using subprocess.run to call tools like cc, cl.exe, brew, and moon.\n- [DYNAMIC_EXECUTION]: The scripts/run-asan.py script performs runtime compilation of C code and replaces the core libmoonbitrun.o library object in the MoonBit installation directory (~/.moon/lib/). While intended for memory validation, this modification of local toolchain artifacts is highly intrusive and could impact environment stability.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 09:04 AM