The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill documents how to write scripts that process untrusted data from various external sources, creating a potential surface for indirect prompt injection. * Ingestion points: Data enters the context through hooks such as handleRequest (HTTP headers/body), contentBasedFlowRouter (AS2 messages), and pipeline hooks like preSavePage, filter, or transform (data records) as described in SKILL.md. * Boundary markers: The documentation does not specify requirements for using delimiters or 'ignore embedded instructions' warnings to isolate processed data from the agent's instructions. * Capability inventory: Scripts have access to the integrator-api for network operations and internal Celigo API calls, and they can perform arbitrary data transformations. * Sanitization: There are no explicit instructions for validating or sanitizing external content before it is processed by the script logic.
[DYNAMIC_EXECUTION]: The primary purpose of the skill is to enable the creation, update, and execution of JavaScript code snippets that run at specific hook points in the Celigo integration pipeline.
[COMMAND_EXECUTION]: The skill documents several celigo CLI commands for managing scripts, including listing, creating, updating, and viewing debug logs in SKILL.md.

writing-scripts