16 tools with basic schemas and descriptions, but significant gaps in parameter documentation and error handling. Most tools have descriptions (10-150 chars, below the 194-char baseline), and input schemas are present but sparse. Parameters lack detailed constraints, enums, and format guidance. No output schemas documented. Error responses are generic JSON without recovery guidance. Tool naming is reasonable (verb_noun pattern mostly followed), but composition issues exist: promote/deploy_to_prod overlap; story_set_context modifies state without clear side-effect documentation. No tool annotations (readOnlyHint/destructiveHint) despite clear risk levels. Secrets appear to be injected server-side (good), but no audit logging visible in tool implementations.

• Add output schemas to all tools. Document field names, types, and required fields. E.g., story_list should return {stories: [{id: string, name: string, status: string, pipeline: string}], total: number}.
• Convert free-form string parameters to enums where valid values are known. E.g., promote(environment: enum['UAT-SFP', 'UAT-S', 'PROD']) and ai_ask(agent: enum['plan', 'build', 'test', 'release', 'operate']).
• Enhance error messages with recovery guidance. E.g., 'Salesforce not authenticated. Call auth_status() to check credentials, or run: copado-hx auth login' instead of bare 'Salesforce not authenticated'.
• Add tool annotations (readOnlyHint, destructiveHint, idempotentHint) to all tool definitions. Mark promote/deploy_to_prod as destructiveHint=true; mark story_list/env_list as readOnlyHint=true.
• Consolidate promote() and deploy_to_prod(). Merge into a single promote(environment, approval_code, validate_only) tool with environment as an enum including 'PROD'. Remove deploy_to_prod() to reduce LLM confusion.

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

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