Both tools lack input schemas entirely and have minimal descriptions (under 20 chars). truecopy-verify: 're-check every pinned skill for drift / poisoning' (56 chars, but vague on parameters). truecopy-status: 'show the pinned set and current verification status' (51 chars, similarly vague). No parameter documentation visible in source. No output schema documented. Tool definitions appear inferred from mcp-serve.mjs rather than explicitly registered with full schemas. STDIO transport limits remote accessibility.

• Add explicit JSON Schema for each tool with all parameters typed (string, boolean, object, array) and described. Example: truecopy-verify should declare whether it accepts a --lock parameter, --config, or operates on a default lock file.
• Expand descriptions to 100-200 characters following the pattern: WHAT (re-verify pinned skills against canonical hashes), WHEN (after updating dependencies or detecting drift), WHAT IT RETURNS (verification report with per-skill status: clean, drifted, poisoned).
• Document output schema explicitly. truecopy-status should return { pinned: [{name, hash, verdict}], summary: {total, clean, drifted, poisoned} }. truecopy-verify should return { report: [{name, status, reason}], allowed: Set }.
• Add parameter descriptions for any inputs. If truecopy-verify accepts a --threshold or --strict flag, document it: 'Threshold for drift detection (0-100, default 95). Higher values tolerate minor changes.'
• Implement error handling guidance. If a lock file is missing or corrupted, return 'Lock file not found at /app/truecopy.lock. Initialize with: truecopy-mcp --lock <path>' instead of a raw error.