delphi-role
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes natural language requests to modify agent configurations, creating a surface where malicious instructions could influence role behavior.
- Ingestion points: User input is captured via the
$ARGUMENTSvariable inSKILL.mdto map requests to CLI commands in steps 1 and 3. - Boundary markers: The skill explicitly requires the agent to "Show the user the diff and wait" before performing final changes, serving as a manual human-in-the-loop review.
- Capability inventory: Commands such as
delphi role add,delphi role build, anddelphi doctorare used to change behavior patterns and check for dependencies like external skills or MCP servers. - Sanitization: The skill relies on manual verification rather than automated validation or sanitization of the arguments before they are processed by the tool.
Audit Metadata