The Agent Skills Directory

[DATA_EXPOSURE]: The skill naturally handles sensitive financial documents such as receipts, invoices, and bank statements. However, it implements clear policies for data handling, including source provenance preservation and explicit boundaries for automation. Access to email and filesystem tools is constrained by instructional guidelines to maintain provenance.
[INDIRECT_PROMPT_INJECTION]: As the skill ingests external documents (PDFs, images, emails) which could contain malicious instructions, it is inherently exposed to indirect prompt injection. The skill mitigates this by instructing the agent to route records failing validation to a review queue and requiring human confirmation for high-risk actions like posting to accounting software.
[COMMAND_EXECUTION]: The included Python script (scripts/receipt_summary.py) performs deterministic data summarization using standard libraries. It does not perform unsafe operations, remote code execution, or privilege escalation. It uses standard json.load for parsing input data.
[EXTERNAL_DOWNLOADS]: The skill uses WebFetch to acquire source material, which is consistent with its primary purpose of extracting data from digital receipts and invoices. No suspicious or untrusted domains were identified, and the mentions of 'Receiptor AI' refer to the vendor's own established tooling.

receipt-processing