Security audit
Security checks for vulnerabilities and agentic risk
This documentation-only skill helps an agent use the Coder CLI for workspace and task management, with powerful but disclosed operations that users should invoke carefully.
Install only if you trust the configured Coder deployment and are comfortable giving the agent access through your Coder session token. Review workspace and task names before delete operations, and treat remote SSH commands as capable of changing files or services inside the target workspace.
The README advertises workspace deletion as a feature, but it does not include any warning that deleting a workspace may be destructive or irreversible. For a markdown skill description, operations affecting user data or system state should include a clear warning so users understand the risk before invoking the skill.
The skill description states that it can SSH into workspaces and run commands, but provides no warning that those commands can modify systems, run arbitrary actions, or affect remote data. Markdown descriptions should disclose behaviors that may impact user systems or integrity.
Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.
- **Workspaces**: List, create, start, stop, restart, delete
- **Remote Commands**: SSH into workspaces and run commands
- **AI Tasks**: Create and manage Coder Tasks with Claude Code, Aider, Goose, etc.
## Prerequisites
Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.
- "List my Coder workspaces"
- "Start my dev workspace"
- "Create a task to fix the auth bug"
- "Check status of my running tasks"
- "SSH into backend and run the tests"
The skill documents coder delete <workspace> -y as a ready-to-run command without an explicit warning that deletion is irreversible and may destroy workspace state, files, or environment configuration. In an operational CLI skill, users may copy-paste commands directly, so omitting a safety warning increases the chance of accidental destructive actions.
The documented coder tasks delete <task-name> -y command removes tasks without any warning about loss of task state, logs, outputs, or the inability to reconnect later. Because this skill is specifically about managing AI coding tasks, users are likely to invoke deletion during routine operations, making accidental data loss more plausible.
The setup instructions tell users to place a session token in configuration, but they do not warn that the token is sensitive credential material requiring careful storage and access control. For markdown files, omission of privacy or credential-handling warnings is in scope when the skill behavior involves sensitive data.
No suspicious patterns detected.