ContextVC exposes 22 tools with basic descriptions and parameter schemas, but falls short of production quality. Most tool descriptions are terse (10-50 chars), leaving LLMs without clear selection guidance. Parameter schemas are present but lack depth: no enums for constrained inputs (e.g., 'json' boolean flags appear in 10+ tools but are never explained), no min/max bounds on numeric parameters, and no guidance on parameter relationships or dependencies. Error handling is minimal, no recovery guidance, no categorization of retryable vs fatal errors. Tool naming is reasonable (verb_noun pattern mostly followed), but composition issues exist: 'review_accept' and 'review_reject' are tightly coupled; 'log_event' and 'harvest' both manipulate state but lack idempotency guarantees. Output schemas are not documented, callers cannot predict what fields to expect. Security: no evidence of permission gates, audit trails, or secret injection patterns despite tools like 'doctor' and 'merge' that modify state.

• Expand all tool descriptions to 100-200 chars. Include WHAT the tool does, WHEN to use it (vs similar tools), and WHAT it returns. Example: 'search_objects' should say 'Search knowledge objects by full-text query with optional scope filtering. Returns ranked results with ID, title, type, and confidence score. Use before get_object to find objects by name or content.'
• Document output schemas for all tools. Define the response structure (fields, types, cardinality). Example: 'search_objects returns {results: [{id, type, title, body, status, scope, confidence, rank}], total_count, next_cursor}'.
• Add enums for constrained parameters. Replace 'json: boolean' with explicit enum values or document what 'true' vs 'false' means. Add scope pattern guidance (e.g., 'scope: glob pattern like "src/**" or "tests/*"').
• Add idempotency and error recovery guidance to state-modifying tools. Example: 'log_event appends to ledger (idempotent if event_type + payload are identical). On failure, returns {error, retry_after_seconds, suggested_action}'.
• Implement permission gates and audit logging. Gate 'doctor --fix', 'merge', 'render' behind permission checks. Log all state-modifying calls with user/agent ID, timestamp, parameters, and outcome for compliance.