The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from external sources such as SMS, social media mentions, and meeting transcripts. This creates a surface for indirect prompt injection where malicious instructions embedded in the webhook payload could influence the agent's transformation logic or entity extraction.
Ingestion points: Raw webhook payloads in JSON format from external services are ingested as described in the 'Phases' section of SKILL.md.
Boundary markers: The skill lacks explicit boundary markers or instructions to treat external payload content strictly as data rather than instructions during the transformation and event-processing phases.
Capability inventory: The skill utilizes the put_page, add_timeline_entry, and search tools. It also references command-line style operations gbrain put and gbrain sync for writing and synchronizing data to the underlying system.
Sanitization: While the skill specifies stripping HTML and escaping script tags to prevent XSS in brain pages, these measures do not prevent natural language prompt injection that could subvert the agent's logic.

webhook-transforms