The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill documents the processing of data from external sources like exports and HTTP requests, which creates an indirect prompt injection surface. The documentation mitigates this by identifying specific ingestion points and platform capabilities.
Ingestion points: Data enters through options.data[], options.record, and options.rawBody as detailed in SKILL.md.
Boundary markers: Scripts are executed in a sandboxed environment as specified in references/schemas/request.yml.
Capability inventory: Scripts have access to the integrator-api for running exports, imports, and flows, as well as sjcl for cryptographic operations as described in SKILL.md and references/integrator-api.md.
Sanitization: The skill promotes structured error return patterns to handle record-level issues without failing entire pages.
[DYNAMIC_EXECUTION]: The skill facilitates the creation and wiring of JavaScript code for execution at various pipeline hook points, which is the primary intended function of the Celigo platform.

writing-scripts