AgentSkillsCN

tidewave-tools-usage

对所有Elixir/Phoenix/Ash开发工作至关重要。在处理Elixir代码、Ecto模式、Ash资源、Phoenix应用或Elixir项目中的数据库时调用此技能。提供MCP工具，用于实时代码评估（通过IEx）、即时模块导航、直接执行SQL、模式内省和文档访问。Tidewave为Elixir任务提供了优于bash/grep/read的优秀替代方案。

SKILL.md
--- frontmatter
name: tidewave-tools-usage
description: CRITICAL for ALL Elixir/Phoenix/Ash development work. Invoke when working with Elixir code, Ecto schemas, Ash resources, Phoenix applications, or databases in Elixir projects. Provides MCP tools for live code evaluation (via IEx), instant module navigation, direct SQL execution, schema introspection, and documentation access. Tidewave provides superior alternatives to bash/grep/read for Elixir tasks.

Tidewave MCP Tools - CRITICAL PRIORITY FOR ELIXIR DEVELOPMENT

MANDATORY: Use Tidewave MCP Tools (Skill) as Primary Interface

When working with this Elixir/Phoenix/Ash codebase, ALWAYS prioritize Tidewave MCP tools over traditional file system operations. Tidewave provides deep integration with the Elixir runtime and superior code intelligence.

Tool Usage Hierarchy

1. Code Evaluation - ALWAYS Use Tidewave

NEVER use Bash to run Elixir code! Instead use mcp__tidewave__project_eval:

  • •Test function behavior and debug issues
  • •Explore modules and their functions
  • •Access IEx helpers (e.g., exports(Module), h(Module.function))
  • •Capture IO output
  • •Pass arguments with the arguments parameter
  • •Set custom timeout for long-running operations

Example:

  • •❌ WRONG: bash: mix run -e "IO.inspect(MyModule.function())"
  • •✅ RIGHT: mcp__tidewave__project_eval: code: "IO.inspect(MyModule.function())"

2. Source Code Navigation - Tidewave First

Before using Grep, Glob, or Read for Elixir code:

mcp__tidewave__get_source_location - Find exact file locations instantly

  • •Works with: Module, Module.function, Module.function/arity
  • •Find dependencies: "dep:package_name"
  • •FASTER than grep/glob for known modules

mcp__tidewave__get_docs - Get documentation without reading files

  • •Module docs: "MyModule"
  • •Function docs: "MyModule.function/2"
  • •Callback docs: "c:GenServer.init/1"

Example:

  • •❌ WRONG: grep: pattern: "defmodule Worker"
  • •✅ RIGHT: mcp__tidewave__get_source_location: reference: "MyApp.Worker"

3. Database Operations - Direct SQL Execution

mcp__tidewave__execute_sql_query - Run SQL directly against Ecto repos

  • •Inspect database schema
  • •Query data (limited to 50 rows, use LIMIT/OFFSET for more)
  • •Supports parameterized queries
  • •Auto-detects available repositories
  • •Returns native Elixir data structures

mcp__tidewave__get_ecto_schemas - List all schemas and their locations

  • •ALWAYS use this before searching for schema files
  • •Returns module names with file paths

Example:

  • •❌ WRONG: bash: psql -c "SELECT * FROM users"
  • •✅ RIGHT: mcp__tidewave__execute_sql_query: query: "SELECT * FROM users LIMIT 10"

4. Dependency Documentation

mcp__tidewave__search_package_docs - Search Hex documentation

  • •Searches project dependencies by default
  • •Can target specific packages
  • •Use BEFORE trying to read dependency source code

5. Error Diagnosis

mcp__tidewave__get_logs - Get application logs

  • •Filter with regex patterns
  • •Tail recent entries
  • •Essential for debugging runtime issues

Workflow Patterns

Understanding a Module

  1. •FIRST: mcp__tidewave__get_docs - Get documentation
  2. •THEN: mcp__tidewave__get_source_location - Find the file
  3. •THEN: mcp__tidewave__project_eval with exports(Module) - List functions
  4. •FINALLY: Read the file if needed for implementation details

Testing Code Changes

  1. •ALWAYS: Test with mcp__tidewave__project_eval before writing
  2. •Example: code: "MyModule.new_function(:test_input) |> IO.inspect()"
  3. •Verify behavior matches expectations
  4. •Only then modify the actual file

Database Work

  1. •START: mcp__tidewave__get_ecto_schemas - Understand data models
  2. •EXPLORE: mcp__tidewave__execute_sql_query - Inspect actual data
  3. •TEST: mcp__tidewave__project_eval - Test Ecto queries
  4. •IMPLEMENT: Make schema/migration changes

Debugging Issues

  1. •CHECK: mcp__tidewave__get_logs - Recent errors
  2. •LOCATE: mcp__tidewave__get_source_location - Find problem code
  3. •TEST: mcp__tidewave__project_eval - Reproduce issue
  4. •FIX: Edit the file with the solution

IEx Helpers Available in project_eval

  • •h(Module) - Get help for a module
  • •exports(Module) - List all exported functions
  • •i(value) - Inspect data structure info
  • •t(Module) - Show types defined in module
  • •b(Module) - Show behaviours module implements
  • •arguments - Access passed arguments array

Database Query Gotchas

When using execute_sql_query:

  • •UUIDs return as 16-byte binaries - cast with ::text (PostgreSQL)
  • •Results limited to 50 rows - use LIMIT/OFFSET for pagination
  • •Use parameterized queries: query: "SELECT * FROM users WHERE id = $1", arguments: [123]

Common Mistakes to Avoid

❌ DON'T:

  • •Use bash to run mix commands for code evaluation
  • •Use grep to find module definitions when you know the module name
  • •Read entire files to find function documentation
  • •Run iex in bash instead of using project_eval
  • •Search file system for Ecto schemas before using get_ecto_schemas

✅ DO:

  • •Use project_eval for ALL Elixir code execution
  • •Use get_source_location for known modules
  • •Use get_docs for documentation
  • •Use get_ecto_schemas first for schema discovery
  • •Use Tidewave MCP tools as your primary interface

Remember: Tidewave Is Your Superpower

The Tidewave MCP server gives you:

  • •Direct access to the running Elixir application
  • •Instant code evaluation with full project context
  • •Database introspection without external tools
  • •Documentation at your fingertips
  • •Source navigation faster than file search

Every time you reach for Bash, Grep, or Read for Elixir code, ask yourself: "Can Tidewave MCP do this better?" The answer is almost always YES.