The Agent Skills Directory

[SAFE]: The skill implements a standard research framework using built-in agent tools (Read, Write, Task, AskUserQuestion) to generate YAML configuration files. All operations are confined to the current project context.
[DATA_VALIDATION]: The included validate_json.py script is a utility for ensuring consistency between generated research results and the defined schema. It uses yaml.safe_load() which is a secure practice for parsing configuration files and prevents unsafe deserialization attacks.
[REMOTE_CODE_EXECUTION]: While the skill uses the Task tool to launch a background research agent, this is a standard platform capability. The prompt for the subagent is transparently documented and does not contain shell escape sequences or remote execution triggers.
[DATA_EXPOSURE]: The skill uses the Read tool to ingest user-supplied field definitions. This is gated by an AskUserQuestion interaction, ensuring the user is aware of and approves the file access.

research-outline