Agent Curator Skill for Claude
| Deep curation of a single agent definition through structured human conversation. Coordinates knowledge research, persona refinement, tooling optimization, and instruction quality improvement. One agent at a time, with due care.
Install
one command, takes just this skill from the repositorynpx skills add https://github.com/majiayu000/claude-skill-registry --skill agent-curatorWhat comes with it
710 bytes besides the instructionThe instruction itself
23 sections, as written by the authorAgent Curator Skill
Activation
Invoke for ad-hoc agent curation or project preparation:
[ACTIVATE:AGENT_CURATOR_V1]
Parameters:
agent: path/to/agent.md
mode: broad | project
project: {project-name} # Required if mode=project
Purpose
Drive a structured conversation to deeply improve an agent definition across all dimensions:
- Knowledge Sources (P0) — Find and integrate esteemed references
- Persona & Doctrine (P1) — Sharpen identity and mental model
- Tooling & MCP (P2) — Optimize tool and server configuration
- Local Corpora (P3) — Materialize high-value knowledge locally
- Instruction Quality (P4) — Refine priorities, modes, and behaviors
Two Modes
Broad Improvement Mode
For general agent enhancement:
- Changes should improve effectiveness across all use cases
- Improvements are candidates for upstream contribution
- Long-term maintenance burden considered
Project-Specific Mode
For preparing an agent for a specific project:
- Changes may include project-specific knowledge/tooling
- Consider whether changes should stay local or go upstream
- Immediate utility prioritized, but long-term impact considered
Execution Flow
Phase 0: Scope Establishment
Present scope selection:
╔═══════════════════════════════════════════════════════════════════════╗
║ AGENT CURATION SESSION ║
╠═══════════════════════════════════════════════════════════════════════╣
║ ║
║ Agent: {agent-name} ║
║ Tier: {focused | expert | phd} ║
║ Current File: {path} ║
║ ║
║ What is the goal of this curation session? ║
║ ║
║ [B] Broad Improvement ║
║ Enhance general effectiveness across all use cases ║
║ Changes are candidates for upstream contribution ║
║ ║
║ [P] Project-Specific Preparation ║
║ Optimize for: {project-name if provided} ║
║ May include project-specific knowledge/tooling ║
║ ║
╚═══════════════════════════════════════════════════════════════════════╝
Wait for human selection before proceeding.
Phase 1: Initial Analysis
Load and analyze the agent definition:
╔═══════════════════════════════════════════════════════════════════════╗
║ INITIAL ASSESSMENT ║
╠═══════════════════════════════════════════════════════════════════════╣
║ ║
║ Agent: {agent-name} ║
║ Mode: {Broad | Project-Specific} ║
║ ║
║ ┌────────────────────────────────────────────────────────────────┐ ║
║ │ DIMENSION SCORES │ ║
║ ├────────────────────────────────────────────────────────────────┤ ║
║ │ Knowledge Sources [██░░░] 2/5 HIGH gap │ ║
║ │ Persona & Doctrine [████░] 4/5 LOW gap │ ║
║ │ Tooling & MCP [███░░] 3/5 MEDIUM gap │ ║
║ │ Local Corpora [█░░░░] 1/5 HIGH gap │ ║
║ │ Instruction Quality [████░] 4/5 LOW gap │ ║
║ └────────────────────────────────────────────────────────────────┘ ║
║ ║
║ Key Observations: ║
║ • {observation-1} ║
║ • {observation-2} ║
║ • {observation-3} ║
║ ║
║ Recommended Focus: Knowledge Sources, Local Corpora ║
║ ║
╠═══════════════════════════════════════════════════════════════════════╣
║ [C] Continue with full curation ║
║ [F] Focus on high-gap dimensions only ║
║ [S] Skip to specific phase ║
╚═══════════════════════════════════════════════════════════════════════╝
Phase 2: Knowledge Sources (P0)
Step 2a: Current State Audit
╔═══════════════════════════════════════════════════════════════════════╗
║ KNOWLEDGE SOURCES — CURRENT STATE ║
╠═══════════════════════════════════════════════════════════════════════╣
║ ║
║ ┌────────────────────────────────────────────────────────────────┐ ║
║ │ EXISTING SOURCES │ ║
║ ├────────────────────────────────────────────────────────────────┤ ║
║ │ ✓ {url-1} │ ║
║ │ Authority: Official | Status: 200 OK | Unique Value: Yes │ ║
║ │ │ ║
║ │ ⚠ {url-2} │ ║
║ │ Authority: Community | Status: 200 OK | Unique Value: ? │ ║
║ │ │ ║
║ │ ✗ {url-3} │ ║
║ │ Authority: Vendor | Status: 404 | Action: Remove │ ║
║ └────────────────────────────────────────────────────────────────┘ ║
║ ║
║ Issues Found: ║
║ • 1 broken URL ║
║ • 2 sources with unclear unique value ║
║ • Missing: authoritative specification references ║
║ ║
╠═══════════════════════════════════════════════════════════════════════╣
║ [R] Research new sources ║
║ [A] Address issues (remove broken, evaluate unclear) ║
║ [N] Next phase (no changes) ║
╚═══════════════════════════════════════════════════════════════════════╝
Step 2b: Knowledge Research
Delegate to agent-knowledge-researcher and web-researcher:
Researching authoritative sources for: {agent-domain}
Using Opus to identify esteemed references...
Using Firecrawl for parallel discovery...
Evaluating:
• Official documentation sites
• Specification repositories
• Academic papers (if applicable)
• Vendor documentation
Step 2c: Source-by-Source Adjudication
Present EACH source individually for human decision:
╔═══════════════════════════════════════════════════════════════════════╗
║ SOURCE ADJUDICATION (1 of N) ║
╠═══════════════════════════════════════════════════════════════════════╣
║ ║
║ Source: {source-title} ║
║ URL: {url} ║
║ Authority: {tier} ({tier-name}) ║
║ ║
║ Unique Value: ║
║ {explanation of what this provides that nothing else does} ║
║ ║
║ Materialization Recommendation: {URL | Local Excerpt | Embed} ║
║ Rationale: {why this materialization} ║
║ ║
╠═══════════════════════════════════════════════════════════════════════╣
║ [A] Add with recommended materialization ║
║ [M] Add with different materialization ║
║ [S] Skip this source ║
║ [D] Show more details ║
╚═══════════════════════════════════════════════════════════════════════╝
Repeat for each source. Never bulk-add.
Phase 3: Persona & Doctrine (P1)
╔═══════════════════════════════════════════════════════════════════════╗
║ PERSONA & DOCTRINE ║
╠═══════════════════════════════════════════════════════════════════════╣
║ ║
║ ┌────────────────────────────────────────────────────────────────┐ ║
║ │ CURRENT INTERPRETIVE LENS │ ║
║ ├────────────────────────────────────────────────────────────────┤ ║
║ │ {current lens text} │ ║
║ └────────────────────────────────────────────────────────────────┘ ║
║ ║
║ Assessment: {Good | Needs sharpening | Missing} ║
║ ║
║ ┌────────────────────────────────────────────────────────────────┐ ║
║ │ PROPOSED IMPROVEMENT │ ║
║ ├────────────────────────────────────────────────────────────────┤ ║
║ │ {proposed improved lens} │ ║
║ └────────────────────────────────────────────────────────────────┘ ║
║ ║
║ Rationale: {why this is better} ║
║ ║
╠═══════════════════════════════════════════════════════════════════════╣
║ [A] Accept improvement ║
║ [K] Keep current ║
║ [E] Edit proposal ║
╚═══════════════════════════════════════════════════════════════════════╝
Similarly for vocabulary calibration and core principles.
Phase 4: Tooling & MCP (P2)
╔═══════════════════════════════════════════════════════════════════════╗
║ TOOLING & MCP ║
╠═══════════════════════════════════════════════════════════════════════╣
║ ║
║ Current Tool Modes: ║
║ audit: {tools} ║
║ solution: {tools} ║
║ research: {tools} ║
║ ║
║ Current MCP Servers: ║
║ • {server-1}: {purpose} ║
║ • {server-2}: {purpose} ║
║ ║
║ Recommendations: ║
║ [+] Add MCP server: {server} ║
║ Rationale: {why this would help} ║
║ ║
║ [~] Modify tool mode: {mode} ║
║ Proposal: {change} ║
║ ║
╠═══════════════════════════════════════════════════════════════════════╣
║ Review each recommendation? [Y/n] ║
╚═══════════════════════════════════════════════════════════════════════╝
Phase 5: Local Corpora (P3)
╔═══════════════════════════════════════════════════════════════════════╗
║ LOCAL CORPORA ║
╠═══════════════════════════════════════════════════════════════════════╣
║ ║
║ Materialization Candidates: ║
║ ║
║ 1. {knowledge-topic} ║
║ From: {source} ║
║ Size: ~{token-estimate} tokens ║
║ Stability: {stable | slow-decay | fast-decay} ║
║ Recommendation: {Materialize | Keep as URL} ║
║ ║
║ 2. {knowledge-topic-2} ║
║ ... ║
║ ║
╠═══════════════════════════════════════════════════════════════════════╣
║ [R] Review each candidate ║
║ [A] Accept all recommended ║
║ [N] No local materialization ║
╚═══════════════════════════════════════════════════════════════════════╝
Phase 6: Instruction Quality (P4)
╔═══════════════════════════════════════════════════════════════════════╗
║ INSTRUCTION QUALITY ║
╠═══════════════════════════════════════════════════════════════════════╣
║ ║
║ Findings: ║
║ ║
║ [HIGH] Instruction Interference ║
║ P1.5 conflicts with P2.3 on {topic} ║
║ Proposed resolution: {resolution} ║
║ ║
║ [MEDIUM] Missing Cognitive Mode ║
║ Convergent mode not defined ║
║ Proposed: {mode definition} ║
║ ║
║ [LOW] Escalation Calibration ║
║ Threshold may be too high for this domain ║
║ Proposed: Lower from 0.7 to 0.6 ║
║ ║
╠═══════════════════════════════════════════════════════════════════════╣
║ [R] Review each finding ║
║ [H] Address HIGH severity only ║
║ [N] No instruction changes ║
╚═══════════════════════════════════════════════════════════════════════╝
Phase 7: Synthesis
╔═══════════════════════════════════════════════════════════════════════╗
║ CURATION COMPLETE ║
╠═══════════════════════════════════════════════════════════════════════╣
║ ║
║ Agent: {agent-name} ║
║ Mode: {Broad | Project-Specific} ║
║ ║
║ Changes Summary: ║
║ Knowledge: +3 sources, -1 broken, 2 URLs validated ║
║ Persona: Interpretive lens sharpened ║
║ Tooling: +1 MCP server (firecrawl) ║
║ Corpora: 1 local excerpt added ║
║ Instructions: 1 conflict resolved, 1 mode added ║
║ ║
║ Score Improvement: ║
║ Knowledge: 2/5 → 4/5 ║
║ Persona: 4/5 → 5/5 ║
║ Tooling: 3/5 → 4/5 ║
║ Corpora: 1/5 → 3/5 ║
║ Instructions: 4/5 → 5/5 ║
║ ║
╠═══════════════════════════════════════════════════════════════════════╣
║ [S] Save updated definition ║
║ [R] Review changes before saving ║
║ [C] Cancel (discard all changes) ║
╚═══════════════════════════════════════════════════════════════════════╝
Final: Contribution Offer (Broad Mode Only)
╔═══════════════════════════════════════════════════════════════════════╗
║ CONTRIBUTE UPSTREAM? ║
╠═══════════════════════════════════════════════════════════════════════╣
║ ║
║ These improvements may benefit other users of this agent. ║
║ ║
║ Would you like to contribute to the upstream agents repo? ║
║ ║
║ [Y] Yes - Prepare PR to turbobeest/agents ║
║ [N] No - Keep improvements local only ║
║ ║
╚═══════════════════════════════════════════════════════════════════════╝
Agent Coordination
This skill uses the agent-curation-orchestrator to:
- Drive the structured conversation
- Coordinate with
agent-knowledge-researcherfor source discovery - Coordinate with
web-researcherfor Firecrawl-based research - Apply changes using appropriate tier editor
Signals
| Signal | Meaning |
|--------|---------|
| CURATION_STARTED | Session begun for agent |
| CURATION_PHASE_COMPLETE | Individual phase finished |
| CURATION_COMPLETE | All phases done, agent updated |
| CURATION_CANCELLED | Session cancelled, no changes |
Integration with Dev-System Phases
During project phases, this skill can be invoked to prepare agents:
Phase 6 Pre-Gate: Preparing for specification work
The typescript-pro agent will be used.
Would you like to curate it for this project? [Y/n]
[ACTIVATE:AGENT_CURATOR_V1]
agent: agents/-03-agents/.../typescript-pro.md
mode: project
project: {current-project-name}
Completion
Outputs:
- Updated agent definition file
- Curation session record in
.claude/curation-logs/ - Optional: PR to upstream agents repo
Next: Resume normal operations or continue with phase work.
Other skills for the same job
different authors, same section of the catalogue
Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy.
Guide for creating effective skills. This skill should be used when users want to create a new skill (or update an existing skill) that extends Claude's capabilities with specialized knowledge, workflows, or tool integrations.
Guide for creating effective skills. This skill should be used when users want to create a new skill (or update an existing skill) that extends Claude's capabilities with specialized knowledge, workflows, or tool integrations.
Replace with description of the skill and when Claude should use it.
Use when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies
This skill should be used when the user wants to "create a skill", "add a skill to plugin", "write a new skill", "improve skill description", "organize skill content", or needs guidance on skill structure, progressive disclosure, or skill development best practices for Claude Code plugins.
Helps users discover and install agent skills when they ask questions like "how do I do X", "find a skill for X", "is there a skill that can...", or express interest in extending capabilities. This skill should be used when the user is looking for functionality that might exist as an installable skill.
Use when creating new skills, editing existing skills, or verifying skills work before deployment
How to use it
Take majiayu000/agent-curator from the repository into ~/.claude/skills for personal
use, or into .claude/skills inside a project.
The agent identifies a skill by the name field in its header. Two skills with the
same name cannot sit side by side — one of them will be ignored.