dhdna-profiler
Pass
Audited by Gen Agent Trust Hub on Oct 24, 2026
Risk Level: SAFENO_CODEINDIRECT_PROMPT_INJECTION
Full Analysis
- [NO_CODE]: The skill consists entirely of instructional markdown. It does not contain executable scripts, shell commands, or configuration files that trigger automated external operations.
- [INDIRECT_PROMPT_INJECTION]: As a text-profiling tool, the skill ingests untrusted text provided by users. This represents an indirect prompt injection surface. However, the instructions mitigate this risk by providing a strict 12-dimension rubric and requiring the agent to only use the designated sample.
- Ingestion points: User-provided text for profiling (SKILL.md).
- Boundary markers: Instructions to 'Establish the sample' and use only 'text designated for this request'.
- Capability inventory: 'Read' and 'Write' tools for processing local text and generating reports.
- Sanitization: Not applicable as there is no code, but the workflow requires the agent to quote source text directly and verify observations against it.
- [SAFE]: The skill includes specific privacy boundaries, instructing the agent not to retrieve extra personal material or upload data to external services (like MindBook) without explicit user instruction. All referenced URLs are for documentation and research purposes (Zenodo/DOI).
Audit Metadata