Skip to content

humanlayer.md

Latest commit

 

History

History
235 lines (161 loc) · 5.82 KB

humanlayer.md

File metadata and controls

235 lines (161 loc) · 5.82 KB

HumanLayer Plugin

Category: Development Version: 2025-10-09 Author: HumanLayer

Commands and agents from HumanLayer, optimized for Claude Code workflows but with app-specific stuff removed. This plugin provides a complete development workflow system with specialized slash commands and research agents. See original repository for more details.

Commands

The humanlayer plugin provides 6 slash commands that implement complete development workflows:

/commit

Creates git commits for changes made during your session.

Features:

  • Groups related changes logically
  • Follows repository commit message conventions
  • Never adds Claude attribution or co-author information
  • Presents commit plan before execution

Usage:

/commit

/create_plan

Interactive planning process that researches your codebase and generates detailed implementation plans.

Features:

  • Spawns parallel research agents to understand existing code
  • Asks clarifying questions based on actual codebase findings
  • Generates plans with phases and success criteria (automated & manual)
  • Saves to thoughts/shared/plans/YYYY-MM-DD-description.md

Usage:

# Interactive mode
/create_plan

# With ticket file
/create_plan thoughts/allison/tickets/eng_1234.md

# Deep analysis mode
/create_plan think deeply about thoughts/allison/tickets/eng_1234.md

/implement_plan

Executes approved implementation plans phase-by-phase with verification checkpoints.

Features:

  • Reads and understands complete plan context
  • Implements each phase before moving to next
  • Runs automated verification (tests, linting, builds)
  • Pauses for manual verification before continuing
  • Updates checkboxes in plan file as work progresses

Usage:

/implement_plan thoughts/shared/plans/2025-10-09-my-feature.md

/validate_plan

Validates that an implementation plan was correctly executed by verifying all success criteria.

Features:

  • Compares implementation against plan specifications
  • Runs all automated verification commands
  • Identifies what needs manual testing
  • Generates comprehensive validation report
  • Catches issues before they reach production

Usage:

/validate_plan thoughts/shared/plans/2025-10-09-my-feature.md

/describe_pr

Generates comprehensive pull request descriptions using your repository's template.

Features:

  • Reads PR description template from thoughts/shared/pr_description.md
  • Analyzes full PR diff and commit history
  • Runs verification commands when possible
  • Saves to thoughts/shared/prs/{number}_description.md
  • Updates PR description directly on GitHub

Usage:

/describe_pr

/research_codebase

Conducts comprehensive codebase research by spawning parallel specialized agents.

Features:

  • Spawns multiple agents in parallel for efficient research
  • Documents findings with file:line references
  • Generates research document with metadata
  • Saves to thoughts/shared/research/YYYY-MM-DD-description.md
  • Creates permanent GitHub permalinks when possible

Usage:

/research_codebase
# Then provide your research question when prompted

Agents

The humanlayer plugin provides 6 specialized agents that can be invoked via Claude Code's Task tool. These agents are designed to be documentarians—they describe what exists without suggesting improvements unless explicitly asked.

Codebase Agents

codebase-analyzer

Analyzes implementation details with precise file:line references.

Use when you need to:

  • Understand HOW specific code works
  • Trace data flow through components
  • Document technical implementation details
  • Get surgical precision on code behavior

Tools: Read, Grep, Glob, LS

codebase-locator

Finds WHERE code lives in the codebase.

Use when you need to:

  • Locate files related to a feature
  • Find test files, configs, or documentation
  • Understand code organization
  • Map out component locations

Tools: Grep, Glob, LS

codebase-pattern-finder

Finds similar implementations and usage examples.

Use when you need to:

  • Find existing patterns to model after
  • See how similar features are implemented
  • Locate usage examples
  • Discover testing patterns

Tools: Grep, Glob, Read, LS

Thoughts Directory Agents

thoughts-analyzer

Deep dives on research topics within the thoughts/ directory.

Use when you need to:

  • Extract key decisions from documents
  • Find actionable insights from past research
  • Understand historical context
  • Filter high-value information from noise

Tools: Read, Grep, Glob, LS

thoughts-locator

Discovers relevant documents in the thoughts/ directory.

Use when you need to:

  • Find existing tickets, plans, or research
  • Locate historical decisions
  • Discover related documentation
  • Map out what thoughts exist on a topic

Tools: Grep, Glob, LS

Web Research Agent

web-search-researcher

Performs comprehensive web research using search and fetch.

Use when you need to:

  • Find modern information beyond training data
  • Research APIs, libraries, or frameworks
  • Discover best practices from web sources
  • Gather technical documentation

Tools: WebSearch, WebFetch, TodoWrite, Read, Grep, Glob, LS

Requirements

Thoughts Directory Structure

Several commands expect a thoughts/ directory with this structure:

thoughts/
├── shared/
│   ├── plans/              # Implementation plans
│   ├── research/           # Research documents
│   ├── prs/               # PR descriptions
│   └── pr_description.md  # PR template (for /describe_pr)
├── allison/               # Personal thoughts (user-specific)
│   └── tickets/
└── searchable/            # Hard links for searching