The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses a helper script scripts/resolve-codex.py to identify and validate the Codex companion script. It uses subprocess.run() to execute node <script> setup --json to check the status of the environment (Node.js, Codex CLI, and authentication). This is a standard and safe operational check within the context of a developer tool.
[REMOTE_CODE_EXECUTION]: The skill integrates with external review and implementation engines (Codex and CodeRabbit). These are executed via local CLI commands (node <script> task --write, node <script> review) or standard subagent dispatches within the AI platform. These integrations follow the 'TRUST-SCOPE-RULE' as they target well-known developer services and require local authentication/installation of the respective tools by the user.
[DATA_EXPOSURE]: The skill reads issues, tickets, and PRs from external trackers (GitHub, GitLab, Jira, Linear) using official CLIs or MCP tools. It implements a 'Jira-absent fallback' which explicitly warns the user if authentication is missing and asks for manual input rather than attempting to bypass security or guess data. All sensitive actions (like pushing code or transitioning tickets) are strictly opt-in and handled via established user-authenticated tools.
[PRIVILEGE_ESCALATION]: The skill does not request or use elevated privileges (sudo). It operates within the user's standard git environment and local working directory.
[PERSISTENCE]: No persistence mechanisms (such as crontabs or shell profile modifications) were detected. The skill operates on a per-session basis.

forge