The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from contract records and document attachments, which could contain malicious instructions designed to influence the agent's behavior.
Ingestion points: Data is ingested from clm_contract_doc.content, sys_attachment contents, ast_contract (description and terms fields), and clm_obligation (short_description and clause reference fields).
Boundary markers: The skill lacks explicit instructions or delimiters to isolate untrusted contract text from the agent's core instructions.
Capability inventory: The agent has access to powerful write and execution tools, including SN-Update-Record, SN-Create-Record, and SN-Execute-Background-Script.
Sanitization: There is no evidence of sanitization or filtering of the extracted contract text before it is used to populate records or influence script generation.
[DYNAMIC_EXECUTION]: The skill uses SN-Execute-Background-Script to perform complex data aggregation via server-side JavaScript.
The script provided in Step 5 assembles a JSON metadata object by querying multiple tables (ast_contract, core_company, clm_obligation) and performing date calculations at runtime.
The use of background scripts for data processing represents a dynamic execution pattern that bypasses standard API restrictions and requires high-level platform permissions.
[COMMAND_EXECUTION]: The skill requests access to the Bash native tool and uses the SN-Execute-Background-Script MCP tool, both of which provide high-privilege environment access for script execution.

contract-metadata-extraction