compdf-word-to-pdf
Pass
Audited by Gen Agent Trust Hub on Aug 15, 2026
Risk Level: SAFEPROMPT_INJECTIONDATA_EXFILTRATIONNO_CODE
Full Analysis
- [PROMPT_INJECTION]: The skill processes user-supplied Word documents, creating an attack surface for indirect prompt injection.\n
- Ingestion points: Word documents identified as source files in SKILL.md.\n
- Boundary markers: Absent; there are no instructions to delimit document content from the agent prompt.\n
- Capability inventory: The skill allows the agent to prepare network requests and manage local credential files.\n
- Sanitization: Absent; the skill does not specify filtering or escaping of document content.\n- [DATA_EXFILTRATION]: The skill transmits document data to the ComPDF API (compdf.com) for processing. The instructions require the agent to obtain user confirmation before each upload, mitigating the risk of unauthorized data transfer.\n- [NO_CODE]: No executable scripts or source code files were found within the skill package; its logic is entirely contained within markdown and YAML configuration files.\n- [SAFE]: The skill uses a local file (api_key) for credential storage and includes specific instructions to exclude this file from version control and never log or display the key. It also restricts the agent from searching for credentials in environment variables or other system paths.
Audit Metadata