Adam — Development agent for Claude Code
Self-improvement analyst.
How to install Adam
Installs to ~/.claude/agents/lukaszraczylo-claude-adam-adam.md
mkdir -p ~/.claude/agents && curl -fsSL https://raw.githubusercontent.com/lukaszraczylo/claude-adam/HEAD/agents/adam.md -o ~/.claude/agents/lukaszraczylo-claude-adam-adam.md Restart Claude Code, or start a new session, for it to be picked up.
What Adam does
name: adam description: "Self-improvement analyst. Reads the adam journal and transcript context, clusters observations, scores them against a deterministic rubric, and writes proposal files for new skills, memory entries, agent edits, hook changes, CLAUDE.md edits and soft deletions. Invoked only via the adam-self-improvement skill (/reflect). Not for: editing skills, agents or memory directly (the main thread applies approved proposals), or reviewing code (use code-reviewer)." tools: Read,
Alternatives in Development
- Voice Call Judge — Cheap, deterministic voice-call rubric judge 63 ★
- Skill Verify Auditor — Rubric compliance auditor for skill Self-Verify steps (dispatch marker skill-verify: ) 57 ★
- Sc PM Agent — Self-improvement workflow executor that documents implementations, analyzes mistakes, and maintains knowledge 55 ★
Full documentation available on GitHub
View Source RepositoryRelated Agents
Plan Judge
Independent reviewer that scores an implementation plan against the fixed plan-quality rubric and returns spec
Reflection
Self-improvement engine. Captures learnings, updates memory files, and identifies gaps in agents and context.
Session Reflector
Generate an end-of-session reflection by reviewing what happened — ops log, thread changes, new observations —
Diverge Critic
Critic subagent for /diverge — clusters, scores, and ranks generated proposals. DO NOT INVOKE directly outside
Habit Judge
Read-only verifier that scores a session, a diff, or a transcript against a habit set and returns an evidence-
Trainer Agent
Improvement Proposal Agent of the Governed Multi-Agent Delivery System. Invoked after every evaluation cycle.
Related Skills
Mem Forget
Active forgetting — archive raw journal transcript past its TTL (default 7 days) out of active recall, after s
Analyze Learnings
Analyze learnings and journal entries to identify improvement opportunities.
Skill Optimizer
A Claude Code skill that audits how well an agent followed a skill — reads JSONL transcripts, scores 5 dimensi
