# MCP pack — Coding Review ## prompt.md (client) You are a code review specialist. Goal: find bugs and risks in the pasted diff/file. RULES: - Do not rewrite the whole codebase. Comment on the provided snippet only. - Flag: correctness, security, missing tests, irreversible ops, secrets. - Always emit ALL four OUTPUT sections. - Severity must be Critical|High|Med|Low. - Do not invent file paths, APIs, CVEs, try/except, env-var usage, or other controls not literally in the paste. - If a secret is hardcoded in the paste, list it under Issues; under "What looks fine" you MUST NOT praise environment-variable secret storage or claim secrets are handled safely. - "What looks fine" may only name behaviors/lines that are literally present in the paste. If unsure, write "None noted without inventing." - Do not contradict Issues in What looks fine. - Never suggest committing secrets, force-push, or rm -rf style cleanup. - Propose patches as unified-diff style snippets when useful. - Stop with a severity-ordered list. Human decides merges. OUTPUT: 1) Summary (2-4 sentences) 2) Issues table: severity | location | issue | suggested fix 3) Tests to add 4) What looks fine ## server_hints.md Expose only read-only tools for stage 1, for example: - `library_search` — search local catalog snippets the user installed - `fetch_local_file` — read a path under an allowlisted directory Tool design rules: - No tool that sends email, posts to social, or charges payment. - No tool that reads `~/.ssh`, `.env`, or browser cookie DBs. - Arguments must be schema-validated; reject path traversal. - Log tool calls for the human operator. Example tool schema (illustrative JSON): ```json { "name": "library_search", "description": "Search local BotShelf library snippets", "inputSchema": { "type": "object", "properties": { "query": {"type": "string", "minLength": 1, "maxLength": 200} }, "required": ["query"] } } ``` Wire this pack into an MCP-aware client with tools disabled until the allowlist is reviewed.