The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted content from document attachments (PDF, PNG, JPG, TIFF), creating a surface for indirect prompt injection where malicious instructions embedded in documents could attempt to influence agent behavior.
Ingestion points: Data is ingested from the sys_attachment table and processed through sn_doc_intelligence_extraction_result records as described in SKILL.md.
Boundary markers: The skill suggests using confidence thresholds and manual review steps to validate output, but does not provide explicit instructions to the agent to ignore embedded commands within the source documents.
Capability inventory: The skill possesses the ability to create/update records (SN-Create-Record, SN-Update-Record) and execute network requests via Bash (curl), allowing processed data to impact the ServiceNow instance.
Sanitization: The procedure includes the use of validation_regex in field mappings to ensure extracted data matches expected formats.
[COMMAND_EXECUTION]: The skill uses the native Bash tool to perform batch operations. Specifically, it provides examples of using curl in a loop to submit multiple document attachments to the ServiceNow REST API.
[DATA_EXFILTRATION]: The skill performs network operations using curl and various REST endpoints. These operations target ServiceNow instances ([instance].service-now.com), which is a well-known service, and are consistent with the skill's stated purpose of document extraction and record management.

document-extraction