The Agent Skills Directory

[SAFE]: The skill uses dynamic context injection in SKILL.md via the command date +%Y-%m-%d. This is a benign operation used solely to scope research recency and establish report timestamps.
[SAFE]: The skill identifies and uses standard developer tools such as pandoc and md-to-pdf (Node.js) for report generation. These are listed in the allowed-tools frontmatter and used for their intended purpose of format conversion.
[SAFE]: The skill defines a research loop that ingests external web data using WebSearch and WebFetch. While this creates an indirect prompt injection surface, the skill implements strong defensive practices: it mandates multi-source validation for critical claims, requires inline citations for every assertion, and instructs the agent to flag source conflicts and low-confidence data explicitly. This reduces the risk of the agent adopting instructions or false data from a single malicious external source.

deep-research