The Agent Skills Directory

[SAFE]: The skill is a management and orchestration layer that relies on standard research tools. It promotes high standards for evidence gathering and data integrity through structured templates and verification gates.\n- [COMMAND_EXECUTION]: The skill contains several utility scripts in Bash and Ruby intended for managing the research project's filesystem. These include commands for directory creation, link integrity checking, and file count auditing. All scripts are scoped to operate within the project's corpus root and are benign in nature.\n- [EXTERNAL_DOWNLOADS]: The skill manages external data acquisition by delegating to a specialized 'run-research' skill. It defines strict rules for these operations, such as requiring verbatim citations and preferring scraped content over search snippets, which reduces the risk of incorrect or malicious data ingestion.\n- [PROMPT_INJECTION]: The skill defines a set of 'Hard rules' and 'Self-correction triggers' to ensure the agent adheres to the research methodology. These are internal constraints for task performance rather than attempts to bypass security filters.

run-corpus-research