FigJam + Terminal — spatial canvas for Claude Code sessions. An Electron desktop application providing a visual workspace for managing multiple terminal sessions, previews, and collaborative development.
Cove is an Electron desktop app with 11 tools spanning PTY management, MCP config, app info, and window control. Tools have basic descriptions and input schemas, but lack depth in parameter documentation, output schema specification, and error handling guidance. Most descriptions are under 100 chars and omit context on when/why to use each tool. Parameters lack constraints (enums, ranges, patterns). No documented output schemas. Error responses are minimal (e.g., 'max_sessions', 'path_not_found') without recovery guidance. Security measures exist (path traversal checks, sessionId validation) but are not reflected in tool descriptions. Overall, definitions are functional but fall short of production-grade agent tool standards.
Get the user's home directory path.
Discover Claude projects from the user's home directory. Reads ~/.claude/history.jsonl for recent projects and scans home directory for directories with .git, package.json, or Python project markers.
Synchronize connected projects by writing a connected-projects.md file to each project's .claude directory. Validates paths to prevent traversal attacks.
Write MCP configuration JSON to a temporary file for a session. Validates session ID and JSON format, prevents path traversal attacks.
Create a new pseudo-terminal (PTY) session with specified working directory and dimensions. Spawns a login shell (/bin/zsh) with proper environment configuration.
Terminate a PTY session. Sends SIGINT first, then SIGKILL after 500ms if needed.
Output schemas not documented. No tool returns a documented schema, LLMs cannot predict what fields to expect or plan downstream calls. E.g., pty:create returns {pid, error?} but this is inferred, not declared.
Parameter constraints missing. Numeric params (cols, rows) lack min/max in descriptions; sessionId validation rules (alphanumeric, dash, 1-128 chars) are enforced in code but not documented in schema. LLMs cannot self-validate.
Error handling lacks recovery guidance. Errors like 'max_sessions', 'path_not_found', 'invalid_session_id' are returned without actionable next steps. E.g., 'max_sessions' should suggest 'Kill an existing PTY with pty:kill() first.'
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 48 | 2026-07-28+ | v2 |
Resize an active PTY session to new dimensions.
Write data to an active PTY session. Sends input to the shell process.
Close the application window.
Toggle maximize/restore state of the application window.
Minimize the application window.
Descriptions too brief and lack context. 'Write data to an active PTY session' (50 chars) does not explain when to use pty:write vs pty:create, or what happens if the session doesn't exist. Baseline is 194 chars.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). Tools like pty:kill and win:close are destructive but lack annotations. Agents cannot distinguish safe from risky operations.