Five tools with basic schemas and descriptions, but significant gaps in parameter documentation and output schema clarity. All tools are READ_ONLY except drift_set_profile (WRITE). Descriptions are present but brief (10-60 chars), below the 50-200 char LLM-optimized baseline. Parameter descriptions exist but lack detail on constraints, formats, and valid ranges. No output schemas documented. Error handling and recovery guidance absent. Tool names follow verb_noun convention (drift_*) but are domain-specific rather than action-verb-first (e.g., 'get_drift_status' would be clearer). Graceful degradation noted in code comments but not reflected in tool descriptions.

• Add output schemas to all tools. Document the structure of drift_status (fields: session_id, timestamp, score, verdict, profile), drift_recent (array of {ts, score, verdict, profile}), drift_explain (result: {score, verdict, explanation}), drift_list_profiles (array of {name, description, threshold, is_active}), and drift_set_profile (result: {success, active_profile}).
• Expand tool descriptions to 80-150 chars. E.g., drift_status: 'Get a summary of the current or specified session, including latest score, verdict, and active profile. Use this to check drift status before scoring new text.' drift_explain: 'Score arbitrary text against a profile without storing results. Use this to test profiles or score one-off text before committing to a session.'
• Add parameter constraints: limit (integer, 1-50, default 10), profile (enum of available profile names, or omit for active), text (string, 1-10000 chars), name (string, required, must match existing profile).
• Document graceful degradation in tool descriptions. E.g., 'If no session exists, returns empty results rather than an error. Call drift_list_profiles first to verify profiles are available.'
• Add error guidance to drift_set_profile: 'Modifies the active-profile file. If the profile name does not exist, returns an error with available profiles. This change persists across sessions.'