Educational framework demonstrating AI agent patterns: agent loop, tool use, permissions, hooks, task management, and subagents. Includes progressive examples (s01-s06) showing how to build coding agents with the Anthropic API.
Agent Blueprint defines 7 tools with basic schemas and descriptions, but falls short of production quality. Tool names lack consistent verb-noun patterns (bash, read_file, write_file are good; glob, todo_write, task are weak). Descriptions are present but generic (10-50 chars for most), missing context on WHEN to use each tool and what happens on failure. Parameter descriptions exist but lack constraints (no enums for status values, no min/max for limits). Error handling is minimal, implementations return plain strings without recovery guidance. Security concerns: bash tool blocks only 4 dangerous patterns (insufficient); no permission gates; no audit logging. Output schemas are undocumented. The todo_write tool schema in code differs from the description provided (items vs todos array). Overall, this reads like a template/reference implementation rather than a production-ready MCP server.
Run a shell command.
Replace exact text in a file once.
Find files matching a glob pattern.
Read file contents.
Run a subagent with fresh conversation context and return its final answer.
Create and manage a task list for your current coding session.
Write content to a file.
Weak tool naming: 'glob' and 'todo_write' lack action verbs. Should be 'find_files' and 'update_todos'. LLMs rely on verb-noun patterns to infer intent.
Descriptions are too generic and lack context. 'Run a shell command' (bash) and 'Find files matching a glob pattern' (glob) don't explain WHEN to use them or what happens on error. Baseline: 194 chars avg; these are 30-50 chars.
No output schemas documented. LLMs cannot plan downstream calls or extract structured data. All tools return plain strings; no field definitions provided.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 50 | 2026-07-28+ | v2 |
Insufficient error handling. bash tool blocks only 4 patterns (rm -rf, sudo, shutdown, /dev/); many dangerous commands slip through (dd, mkfs, curl to exfil). No recovery guidance in error messages.
Schema mismatch: todo_write definition in code uses 'items' array with 'activeForm' field, but provided schema uses 'todos' array without 'activeForm'. Inconsistency will cause tool invocation failures.