The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill documents mechanisms for processing external data via background task payloads and AI tool integrations.
Ingestion points: Untrusted data enters the agent context through task payload arguments, metadata fields, and ai.tool inputs defined in core-reference.md and advanced-reference.md.
Boundary markers: While explicit delimiters for prompt interpolation aren't specified in the patterns, the instructions focus on structured TypeScript code development.
Capability inventory: The skill enables capabilities including network requests (retry.fetch in core-reference.md), system package management (aptGet in config-reference.md), and script execution (python.runScript in advanced-reference.md).
Sanitization: The skill mitigates risks by promoting schemaTask with zod in SKILL.md and core-reference.md for strict runtime validation and type-safe processing of all incoming data.
[DYNAMIC_EXECUTION]: The skill documents the use of @trigger.dev/python for executing inline Python code and script files (python.runInline, python.runScript). These are standard features of the Trigger.dev platform for processing data in background workers.
[EXTERNAL_DOWNLOADS]: The instructions reference the installation of official packages from the @trigger.dev scope via npm and link to official documentation at trigger.dev. These references target a well-known service and are documented neutrally.

trigger-dev