referral-abuse-guardrails
Pass
Audited by Gen Agent Trust Hub on Oct 26, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill consists of instructional markdown and reference files providing business logic for fraud prevention. It contains no executable scripts, shell commands, or network-active components.
- [INDIRECT_PROMPT_INJECTION]: The skill has a data ingestion surface through the 'Interview' workflow where it processes user-provided details about referral programs. However, this is for generating a text-based specification for human review, and the skill lacks any capabilities (network, file-write, or code execution) that could be exploited via injection.
- Ingestion points: User input during the 'Interview' process defined in
SKILL.md. - Boundary markers: Not explicitly defined in the agent instructions.
- Capability inventory: The skill is limited to text generation; it has no file system write access, network operations, or subprocess execution capabilities.
- Sanitization: None implemented for the ingested data.
- [SAFE]: References to other skills (e.g.,
mbfinotti/partnerships-skills@...) are to resources owned by the same author, representing normal modular skill design rather than unverifiable external dependencies.
Audit Metadata