The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes several bundled shell scripts (generate-project-slug.sh, update-industry-metadata.sh, finalize-candidates.sh, prepare-phase3-data.sh) to manage project initialization, metadata updates, and data formatting. These scripts use standard tools like jq and python3 for local processing.
[EXTERNAL_DOWNLOADS]: Research is conducted via authorized search tools and numerous reputable external APIs, including OpenAlex, Semantic Scholar, arXiv, PubMed, USPTO, Lens.org, EPO OPS, EUR-Lex, SEC EDGAR, and FDA Open Data. These are well-known technology and academic services used for their intended purpose.
[DATA_EXFILTRATION]: No unauthorized data exfiltration patterns were detected. Research data and configuration files are stored locally within the project's directory structure in the user's workspace.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external web and API sources to generate trend candidates. While this represents a data ingestion surface, the risk is mitigated by the specialized nature of the subagents used (e.g., trend-web-researcher) and the inclusion of a dedicated candidate review phase (Phase 2.5) to verify signal quality and distinctiveness.
[SAFE]: The skill demonstrates safe operational practices, including configuration disclosure to the user before research begins, the use of semantic project identifiers, and robust logging of its internal phases.

trend-scout