The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local Node.js script located at scripts/validate_capture.mjs to verify the integrity and structure of input manifest files before proceeding with task creation.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from manifest.json and content.md files which is then interpolated into prompts for sub-tasks, creating a potential multi-step injection chain.
Ingestion points: Data is ingested from manifest.json and content.md via absolute file paths provided to the script and task briefs.
Boundary markers: Instructions in SKILL.md and references/thread-brief-contract.md tell the agent to use the source only as 'art direction' and art evidence, but these natural language boundaries do not use strict delimiters to separate untrusted data from instructions.
Capability inventory: The sub-tasks created by this skill have extensive capabilities including file system access, version control operations (commit), and site deployment via the @Sites plugin.
Sanitization: There is no evidence of sanitization, filtering, or escaping for the strings extracted from the manifest (such as title or prompt) before they are placed into the Thread brief contract.

build-daily-inspiration-sites