Animus has 10 tools with mixed quality. Naming is generally clear (verb-noun pattern: list, update, resolve, record). However, critical gaps exist: 7 of 10 tools lack documented output schemas (listTools, updatePermission, updateGroupPermission, listApprovals, resolveApproval, dismissTrustRamp, onApprovalRequest, onApprovalResolved). The two cognitive tools (recordThought, recordCognitiveState) have extremely detailed input schemas but no output documentation. Descriptions are present but often generic (e.g., 'List all tools with their permission settings' is 40 chars, below the 50-char baseline for clarity). Parameter descriptions exist but lack constraint details (e.g., 'mode' enum values are listed but no guidance on when to use each). Error handling is not visible in the source. No evidence of permission gates, audit trails, or recovery guidance.

• Document output schemas for all 10 tools. For listTools, specify: {tools: [{name: string, source: string, mode: 'off'|'ask'|'always_allow', trustRampDismissed: boolean}], total: number}. For subscription tools, clarify event structure and subscription lifecycle.
• Expand tool descriptions to 80-150 chars. Example: 'List all available tools and their current permission modes (off, ask, always_allow). Use this to audit tool access before updating permissions.' This guides LLM selection.
• Add constraint details to parameter descriptions. For 'mode' enum, explain: 'off = tool disabled, ask = require approval per call, always_allow = auto-approve all calls from this agent.'
• Add error handling guidance to WRITE tools. Example for updatePermission: 'If tool not found, call listTools() first to verify the tool name. If permission denied, the agent lacks admin scope.'
• Declare permission scopes for each tool. Example: updatePermission requires scope 'admin:tools'. Add to tool description: 'Requires: admin:tools scope.'