The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/scaffold_connector.py script uses subprocess.run to execute forge create and npm install. This is standard behavior for a development scaffolding tool. The script uses argument lists instead of raw shell strings, which mitigates command injection risks. The tools being executed (forge and npm) are well-known developer utilities.
[CREDENTIALS_UNSAFE]: The skill explicitly includes security rules in SKILL.md (Rule 2) instructing the agent to never ask for credentials in chat and to direct users to use forge login in their own terminal. This is a positive security practice.
[EXTERNAL_DOWNLOADS]: The skill instructions and scripts refer to official Atlassian npm packages such as @forge/teamwork-graph, @forge/api, and @forge/kvs. These are legitimate dependencies for the intended development tasks.
[DATA_EXFILTRATION]: No patterns of data exfiltration or unauthorized network access were detected. The script extracts the domain from the user-provided API URL solely to populate an egress allowlist in the generated manifest.yml file, which is a security requirement for Forge apps.

forge-connector