payabli-bills
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external, untrusted data in the form of vendor invoices and OCR results. This represents a potential indirect prompt injection surface if the invoices contain malicious instructions for the AI.
- Ingestion points: SKILL.md (OCR data via
POST /Import/ocrDocumentFormandPOST /Import/ocrDocumentJson). - Boundary markers: None explicitly specified for prompt delimitation, but functional instructions serve as a boundary.
- Capability inventory: File system read access (
payabli-integration.md) and network API calls to Payabli endpoints. - Sanitization: The skill provides a mitigation by instructing the agent to explicitly map specific extracted fields (
billNumber,netAmount, etc.) to a new payload rather than spreading the raw data object, which helps prevent data or instruction bleeding from the OCR response into API calls. - [SAFE]: All external references and documentation links (
docs.payabli.com) point to the official infrastructure of the skill's author ('payabli'). - [SAFE]: The skill follows standard configuration patterns, such as reading a local integration file (
payabli-integration.md) for SDK settings. - [SAFE]: No obfuscation, hardcoded credentials, unauthorized persistence mechanisms, or remote code execution patterns were identified.
Audit Metadata