The Agent Skills Directory

[SAFE]: The skill facilitates research by ingesting untrusted external data, which introduces a potential surface for indirect prompt injection. 1. Ingestion points: Data enters the agent's context through seed URLs and search tool results during the research loop (Node D). 2. Boundary markers: The instructions do not define specific delimiters for separating external content from the agent's system instructions. 3. Capability inventory: The skill is designed to perform network read operations for research and file system writes to generate ledgers and reports within a user-defined output directory. 4. Sanitization: The primary security control is a human-in-the-loop mitigation strategy. The workflow includes mandatory checkpoints where the user reviews all gathered materials (Node G) and the final report draft (Node L) before completion. This ensures that any malicious instructions hidden in external sources are subject to human oversight rather than being executed autonomously.

loom-enhanced-research