The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill describes building tasks that process payloads from external sources (e.g., backend code or webhooks), which constitutes an ingestion point for untrusted data.
Ingestion points: The run function in SKILL.md and basic-tasks.md accepts arbitrary payload data.
Boundary markers: Documentation does not explicitly specify delimiters for isolating prompt content within payloads, although it emphasizes structural validation.
Capability inventory: Tasks can perform network requests (fetch), database operations, and execute logic with high-tier machine resources.
Sanitization: The skill strongly recommends the use of schemaTask with Zod validation to ensure payloads conform to expected types and structures.
[EXTERNAL_DOWNLOADS]: The config.md file explains how to use build extensions to download and install system packages (via aptGet) and CLI tools (via additionalPackages) during the deployment process. These are documented features of the Trigger.dev platform.
[COMMAND_EXECUTION]: Documentation in config.md describes how to execute shell commands during the build process, such as using chmod to set executable permissions in custom build layers.
[DYNAMIC_EXECUTION]: The config.md file details the use of a Python extension that allows for running inline Python code (python.runInline) or executing standalone Python scripts within tasks.

trigger-dev-tasks