OpenForge exposes 3 tools for skill management with basic schemas and descriptions. All tools have input schemas with type definitions and descriptions present. However, descriptions are minimal (10-30 chars), falling below the 50-200 char baseline for LLM-optimized tools. Parameter descriptions lack actionable detail (e.g., 'source' has no guidance on valid formats or examples). No output schemas are documented. Error handling is absent, no guidance on recovery paths, retryability, or what happens on failure. Tool naming follows verb_noun convention but lacks clarity on mutual exclusivity and side effects. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite skills.remove being destructive.

• Expand tool descriptions to 50-200 chars. Example: 'skills.install: Install one or more skills from a source (URL, registry, or local path). Specify source and optional skill_names to install. Returns installed skill metadata and any errors.'
• Document output schemas for all three tools. Specify fields like skill_id, name, version, status, error_message, and their types.
• Add enum or pattern constraint to 'source' parameter. Example: 'source (string, required): URL (https://...), registry name (e.g., "official"), or local file path (/path/to/skills.json). Min 1 char.'
• Add tool annotations: mark skills.remove with destructiveHint=true; mark skills.search with readOnlyHint=true; mark skills.install with destructiveHint=true (modifies state).
• Add error handling guidance to descriptions. Example: 'If source is invalid, returns error with available sources. If skill not found, suggests similar names. Retryable on network errors.'