The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process data from external Bitrix24 portals (such as CRM leads, deals, and chat messages), creating a surface for indirect prompt injection where malicious instructions could be embedded in the processed data.
Ingestion points: The skill processes API responses in scripts/bitrix24_client.py and event payloads in scripts/offline_sync_worker.py.
Boundary markers: Instructions in SKILL.md explicitly warn the agent to treat Bitrix24 content as untrusted, but the skill does not enforce technical delimiters or explicit safety warnings when interpolating this data into the LLM's context.
Capability inventory: The skill has the ability to perform network requests to Bitrix24 portal domains and write to local audit/state files in the .runtime/ directory using standard file operations and the urllib library.
Sanitization: While the skill implements secrets masking for logging purposes, it does not include filters or validation logic to detect or neutralize natural language instructions hidden within CRM fields or chat messages.

bitrix24-agent