local-llm-bridge
Pass
Audited by Gen Agent Trust Hub on Aug 9, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill processes external file data and user instructions to form prompts for a local LLM, creating an indirect prompt injection surface.\n
- Ingestion points: Input files (INPUT_FILE), persona configurations (PERSONA_FILE), and instruction strings are all ingested into the agent context in SKILL.md examples.\n
- Boundary markers: The documentation lacks specific instructions for the agent to ignore potentially malicious commands embedded within the provided input files.\n
- Capability inventory: The skill has the ability to execute shell commands via Bash and write to the local filesystem (Write), which could be abused if an injection is successful.\n
- Sanitization: No evidence of input sanitization or filtering was found in the skill's operational description.\n- [COMMAND_EXECUTION]: The skill orchestrates tasks by constructing and executing shell commands (e.g., calling ./scripts/run_agent.py). This local command execution is a core part of its functionality but requires proper handling of arguments to prevent local command injection if instructions are not correctly escaped.
Audit Metadata