Daedalus CLI exposes 19 tools with minimal descriptions and incomplete schemas. All tools have basic descriptions (10-50 chars), well below the 50-200 char baseline for LLM-optimized tools. Most parameter descriptions are generic ('File path to read', 'Command to execute'). No output schemas are documented. Critical tools like 'terminal' and 'patch' lack error handling guidance. Tool names are verb-based (good), but descriptions do not explain WHEN to use each tool or what happens on failure. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear risk levels. Composition is poor: 'delegate_task' and 'handoff_task' appear to do the same thing; 'set_context_variable' is a generic state mutation with no constraints.

• Expand all tool descriptions to 50-200 characters. Include WHAT the tool does, WHEN to use it (vs similar tools), and what happens on failure. Example: 'Execute a shell command in the project root. Returns stdout/stderr. Non-zero exit is an error, use error message to guide retry or fallback. Irreversible: use with caution.'
• Document output schemas for all tools. Specify field names, types, and whether results are paginated. Example for search_files: '{results: [{path: string, line: number, context: string}], total: number, hasMore: boolean}'
• Add tool annotations: mark read-only tools with readOnlyHint=true; mark destructive tools (write_file, patch, terminal, generate_image) with destructiveHint=true; mark idempotent tools (get_*, list_*, search_*) with idempotentHint=true.
• Consolidate 'delegate_task' and 'handoff_task' into a single 'handoff_task' tool. If they differ semantically, rename one to clarify the distinction (e.g. 'delegate_to_specialist' vs 'handoff_to_next_agent').
• Add parameter constraints and validation hints. For 'terminal': document shell (bash/sh/cmd), timeout (seconds), working directory default. For 'patch': specify patch format (unified diff, git diff). For 'search_files': clarify pattern syntax (regex, glob, or literal).