Terminal-mcp defines 4 tools with basic schemas and descriptions. All tools have input schemas with typed parameters and descriptions present. However, descriptions are minimal (10-80 chars, well below the 194-char baseline), lack context on WHEN to use each tool, and omit recovery guidance. No output schemas documented. Tools lack error handling guidance, and the exec tool (IRREVERSIBLE risk) has no confirmation/dry-run pattern. Parameter descriptions are terse and lack format/constraint details. No tool annotations (readOnlyHint/destructiveHint) despite clear risk levels. Naming is verb-forward (exec, shell_output, shell_wait_for, shell_snapshot) but lacks clarity on relationships, shell_* tools are not obviously related to session management. No pagination, no batch operations. Audit logging exists server-side but is not exposed to the agent.

• Add output schemas for all tools. Document what fields are returned (e.g., exec returns {exit_code, stdout, stderr, session_id}). This enables LLM planning and chaining.
• Expand tool descriptions to 100 - 200 chars. Include WHEN to use each tool. E.g., 'shell_output: Retrieve accumulated output from a long-lived shell session. Use this after sending a command via exec to check if it has completed. Returns stdout/stderr and session state.'
• Add tool annotations: mark exec with destructiveHint=true; mark shell_output, shell_wait_for, shell_snapshot with readOnlyHint=true. This signals retry safety to the LLM.
• Implement a dry-run or confirmation pattern for exec. E.g., add a 'dry_run' boolean parameter that returns what would execute without actually running it. Or require explicit confirmation before destructive commands.
• Document parameter constraints in descriptions. E.g., 'pattern: Regex or literal string to match in output (max 1000 chars). Use regex syntax for complex patterns.' Add enums for timeout_ms ranges (100 - 60000 ms).

Inferred effective spec: <=2025-11-25.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry