The Agent Skills Directory

[SAFE]: The skill serves as a router for academic research tasks, identifying the user's research stage and directing them to specific sub-skills (e.g., learning, discovery, strategy). It explicitly avoids accessing private workspaces or local sensitive directories unless provided by the user.
[DATA_EXFILTRATION]: No evidence of data exfiltration. The skill uses public-source grounding from reputable domains like equator-network.org, prisma-statement.org, and cochrane.org. There are no network requests to unknown or suspicious domains.
[CREDENTIALS_UNSAFE]: No hardcoded credentials, API keys, or secrets were found. The skill includes a self_check.py script that actively audits for accidental inclusion of local private paths (e.g., /Users/, ~/.claude), which is a security best practice.
[PROMPT_INJECTION]: The skill instructions do not contain any bypass attempts or malicious overrides. The 'ADHD-friendly' mode is an opt-in formatting feature with strict constraints on data persistence and behavior, posing no security risk.
[COMMAND_EXECUTION]: The provided Python script scripts/self_check.py is a benign validation tool used to ensure skill integrity and the absence of hardcoded local dependencies. It does not execute arbitrary remote code.

rw-research-router