Karen — Project Management agent for Claude Code
Verify claims that a task, fix, feature, or project is complete by exercising the real path and comparing observed behavior with the claim.
How to install Karen
Installs to ~/.claude/agents/brew-martin-agent-config-karen.md
mkdir -p ~/.claude/agents && curl -fsSL https://raw.githubusercontent.com/brew-martin/agent-config/HEAD/agents/karen.md -o ~/.claude/agents/brew-martin-agent-config-karen.md Restart Claude Code, or start a new session, for it to be picked up.
What Karen does
name: Karen description: Verify claims that a task, fix, feature, or project is complete by exercising the real path and comparing observed behavior with the claim. Use after work is marked done, when tests pass but real use fails, or when a completion summary needs an independent reality check. color: yellow
Karen
Make "done" mean "works."
Method
- Define the claim. Read the request, acceptance criteria, changed files, and relevant project instructions. Turn the claimed co
Alternatives in Project Management
- Task Checker — Use this agent to verify that tasks marked as 'review' have been properly implemented according to their speci 267 ★
- Grainulator — Autonomous research sprint subagent 87 ★
- Context Gathering — Use when creating a new task OR when starting/switching to a task that lacks a context manifest 64 ★
Full documentation available on GitHub
View Source RepositoryRelated Agents
Workflow Baseline Researcher
Investigates one area of an existing codebase for the project baseline — observed structure, marked inferences
Dev Reader
Reads ONE project document in isolation and returns its claims as JSON, each observable claim verified against
Resection Implementation Reader
Cold reader of a project's actual current implementation — source, tests, config, and observable running behav
Accuser
Prosecution in /objection. Reviews a diff looking for defects that break behavior, with a proof path for each
Plan Worker
Executes plan-graph tasks: fetches the next ready task, claims it, does the work, verifies it and marks it don
Plan Completion Checker
Use this agent when you need to verify whether a project plan generated by the project-planner agent has been
Related Skills
Deepen
Spare-budget design pass. Make one shallow module deep — smaller interface, behavior held, tests green before
De Engineer
De-engineering pass — remove accidental complexity from a target while preserving behavior, tests green after
Proof Of Done
Claude Code plugin that stops a coding agent from ending its turn with 'tests pass', 'build succeeds' or 'depl