hyper-hwp
Pass
Audited by Gen Agent Trust Hub on Nov 19, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes an external binary,
rhwp, to process document files. It uses a bundled Python script (assets/tools/probe_rhwp.py) to verify the binary's presence and version. Command execution is handled viasubprocess.runwith structured argument lists, which avoids shell injection vulnerabilities. The instructions explicitly forbid the silent installation of binaries or dependencies. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted document content, which constitutes an indirect prompt injection surface where malicious instructions could be embedded in document text or metadata.
- Ingestion points: Content from external HWP/HWPX files is read into the agent's context through CLI tools like
export-text,explain, andfields. - Boundary markers: The skill provides comprehensive instructional guardrails in
SKILL.mdandrules/edit-safety.md, explicitly directing the agent to treat all document content as data and to ignore any embedded assistant instructions or system prompts. - Capability inventory: The skill possesses the ability to read, modify, and generate documents. It also includes specific security commands like
rhwp threat-scanandrhwp armorto inspect documents for malicious patterns before processing. - Sanitization: No programmatic sanitization or escaping of document content is implemented; the skill relies on high-level behavioral constraints defined in the instructions.
Audit Metadata