skills/microsoft/hve-core/rai-planner/Gen Agent Trust Hub

rai-planner

Pass

Audited by Gen Agent Trust Hub on Sep 21, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • Command Execution: The skill instructs the agent to execute a local script using npm run rai:sign -- -ProjectSlug {slug} in references/backlog-handoff.md. The {slug} variable is interpolated from the session state, which represents a potential command injection surface if the project slug is not validated.
  • Indirect Prompt Injection: The skill processes user-provided descriptions of AI systems to generate planning artifacts, creating a surface for instructions embedded in that data to influence agent behavior.
  • Ingestion points: User interviews and system descriptions are gathered as described in references/capture-coaching.md.
  • Boundary markers: No specific delimiters or instructions to ignore embedded commands are present in the reference guidance.
  • Capability inventory: The skill performs file system read/write operations for assessment documents and executes shell commands for artifact signing as described in references/backlog-handoff.md.
  • Sanitization: There is no explicit requirement or mention of sanitization or validation of the input data before it is used in documentation or command arguments.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 21, 2026, 08:44 PM
Security Audit — agent-trust-hub — rai-planner