11 tools with basic schemas and descriptions, but significant gaps in parameter documentation and error handling. All tools have names starting with action verbs (login, create, get, delete, like, repost) and descriptions present. However, most parameter descriptions are minimal (10-30 chars), falling below the 72-char baseline. No output schemas documented. Error handling returns JSON strings but lacks recovery guidance or error classification. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear risk levels. Schemas use Zod validation but descriptions do not explain constraints (e.g., 'limit' says '1-100' only in schema, not description text). No mention of prerequisites, dependencies, or when to use each tool vs. alternatives.

• Add tool annotations: mark delete-post with destructiveHint=true, like-post/repost with idempotentHint=true, get-* tools with readOnlyHint=true. This guides agent planning and prevents accidental misuse.
• Expand parameter descriptions to 50-150 chars. E.g., 'uri' → 'The Bluesky post URI (format: at://did:plc:.../app.bsky.feed.post/...). Obtain from get-timeline or get-profile responses.'
• Document output schemas for each tool. E.g., create-post returns {status: 'success'|'error', message: string, uri?: string, cid?: string}. This enables LLMs to chain tools correctly.
• Replace error messages with actionable recovery guidance. E.g., 'Failed to login: invalid credentials. Verify identifier and password in .env or pass them as parameters. If using .env, ensure BLUESKY_IDENTIFIER and BLUESKY_PASSWORD are set.'
• Remove identifier and password from login tool parameters. Require .env-only secret injection. If manual login is needed, create a separate tool that accepts only a session token, not raw credentials.