didit-proof-of-address
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external, untrusted documents such as utility bills and bank statements for OCR extraction and parsing.
- Ingestion points: The
scripts/verify_address.pyscript and documented API endpoints ingest user-provided PDF and image files. - Boundary markers: There are no explicit instructions or delimiters to the LLM to ignore embedded text within the document contents during processing.
- Capability inventory: The skill uses the
requestslibrary to upload these documents to the vendor's verification endpoint. - Sanitization: The script performs standard file type and size validation but does not sanitize the content of the documents before submission.
- [SAFE]: The skill utilizes vendor-specific domains (
didit.me,didit.tech) for all API communications, which is consistent with the author's identity (didit-protocol). - [SAFE]: Authentication is handled via the
DIDIT_API_KEYenvironment variable, aligning with secret management best practices for AI agents.
Audit Metadata