mantis-pipeline-adapter
Pass
Audited by Gen Agent Trust Hub on Jul 15, 2026
Risk Level: SAFE
Full Analysis
- Deterministic Orchestration Guidance: The skill instructs the agent to guide users toward deterministic control flow (code-based) rather than allowing the LLM to control the pipeline execution. This is a positive security practice that limits non-deterministic behavior and reduces the attack surface for prompt-based manipulation of logic.
- Environment Isolation via MCP: It recommends using the Model Context Protocol (MCP) to interface with sensitive environments like VMs or physical hardware. This architectural pattern provides a layer of isolation and controlled API access, which is a recommended security measure for systems performing potentially risky tasks like vulnerability reproduction.
- State Management Practices: The instructions detail managing state via structured local files (e.g.,
workspace/findings/) and rotating logs (e.g.,learnings.jsonl) to prevent data bloat. These operations are standard for the skill's intended purpose of managing security research findings and do not involve unauthorized data access or exfiltration. - Structured Data Handling: The skill emphasizes adherence to strict schemas (e.g.,
schema.json) and using UUIDs for data referencing. These practices help ensure data integrity and reduce the risk of processing malformed or malicious inputs by minimizing the amount of raw text the LLM must process and repeat.
Audit Metadata