persona-plan-marshall-agent
Pass
Audited by Gen Agent Trust Hub on Jul 11, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill implements a robust security policy by forbidding unsafe Bash patterns, including command chaining (&&, ;), shell substitution ($()), and heredocs, which significantly reduces the risk of shell-based injection attacks.
- [SAFE]: It mandates the use of structured tools (Read, Write, Glob, Grep) for file system interactions and explicitly discourages raw shell commands like 'cat' or 'find', ensuring that file operations are executed in a non-prompting and predictable manner.
- [SAFE]: The 'Research Topics' section describes a secure workflow for handling untrusted data from the web by using an isolated, read-only reader context and a deterministic validation gate to sanitize external information before it is consumed by the main agent context.
- [SAFE]: The skill enforces strict dependency governance, requiring agents to obtain explicit user approval before adding any external libraries, frameworks, or tools, preventing unauthorized modifications to the project's dependency graph.
- [SAFE]: It prevents command-line drift and unauthorized execution by requiring agents to follow documented workflows verbatim, use only approved script subcommands, and comply with the declared archetype (e.g., knowledge, workflow) of every loaded skill.
Audit Metadata