skills.rest
watchthelight
watchthelightCOMMUNITY
Explore all skills by this author →
WebsiteGitHub

bmad-agent-builder

Builds, edits, and analyzes persona-driven agent skills through conversational discovery.

Updated Sep 3, 2026
Search Tags:#prompt engineering#skill authoring#agent builder#bmad#persona design#memory architecture#conversational discovery
One-click install
npx skills add https://github.com/watchthelight/shatterfish --skill bmad-agent-builder-watchthelight
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: bmad-agent-builder
Source: https://github.com/watchthelight/shatterfish/tree/main/.claude/skills/bmad-agent-builder
Command: npx skills add https://github.com/watchthelight/shatterfish --skill bmad-agent-builder-watchthelight

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires tiktoken, and includes scripts (resource) and references (resource) and assets (resource) components.

What problem does it solve? Designing a well-structured AI agent skill requires balancing persona design, capability prompts, memory architecture, and quality standards, which is difficult to do consistently by hand. ## Core Features & Use Cases - Create Agents: Guides conversational discovery to build stateless, memory, or autonomous agents with SKILL.md, sanctum templates, and wake scripts. - Edit Agents: Makes targeted changes to existing agents while preserving their design intent and cross-file coherence. - Analyze Agents: Runs quality lenses over an agent and produces an actionable report against the prompt-quality canon. - Use Case: Describe a personal coding coach you want, and the builder discovers whether it needs memory, drafts the persona and capabilities, scaffolds the sanctum, and ships a complete agent directory. ## Quick Start Ask the assistant to create a new agent that acts as a writing editor with persistent memory between sessions.

Frequently Asked Questions about bmad-agent-builder

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I create a custom AI agent skill with persistent memory?▼

Invoke the builder with a description of your agent and answer discovery questions about memory needs. It emits a lean bootloader SKILL.md plus a sanctum of PERSONA, CREED, BOND, MEMORY, and CAPABILITIES templates, with wake.py and init-sanctum.py scripts.

What is the difference between stateless, memory, and autonomous agents?▼

Stateless agents ship everything in one SKILL.md with no memory. Memory agents add a persistent sanctum reloaded each session. Autonomous agents add PULSE.md for scheduled self-directed wake behavior.

Can I edit an existing agent without rebuilding it?▼

Yes, the Edit intent makes targeted changes while preserving the agent's design. It assesses cascade effects across SKILL.md, sanctum templates, and capability references, then validates coherence and runs lint gates afterward.

Does the builder support non-interactive or headless execution?▼

Yes, pass --headless or -H to run the builder non-interactively. It resumes from the memlog automatically and emits a JSON status object on completion instead of an interactive handoff.

How does the builder enforce prompt quality in generated agents?▼

It loads a prompt-quality canon as the governing standard for every capability prompt, applies leanness tests while protecting persona content, and runs scan-path-standards.py and scan-scripts.py lint gates before shipping.

Related Skills

Explore More Software Engineering Skills→
wahabmalikwahabmalik

bmad-agent-builder

Create, modify, and analyze AI Agent Skills through conversational discovery.

Community
Advanced
thevibeworksthevibeworks
28

skill-creator

Create and iteratively improve Claude skills through evaluation loops.

Official
Advanced
xuezhaojunxuezhaojun

custom-registries

Configure private registries in oyoc.json to share hooks, agents, and skills.

Community
Intermediate
scamaiscamai

Agentic UX Design - Relationship-Centric Interfaces

Guide agentic UX design with memory architectures and three-stage trust evolution.

Official
Advanced