mycelium
Evidence-gated product discovery, delivery, and launch discipline
All Skills in This Repository (31)
Pure Emerald Level Indicatorsdiamond-render
Render state diagrams from .claude/diamonds/active.yml using Mermaid stateDiagram-v2.
ping
Verify Mycelium plugin system components and libraries are operational.
define-done
Define outcome-based Definitions of Done for diamond behaviors in the Mycelium framework.
corrections-audit
Analyze correction logs to detect patterns and generate risk mitigation reports.
regulatory-review
Assess EU AI Act applicability and Article 50 transparency obligations.
framework-health
Evaluate Mycelium framework process metrics including cycle velocity, discard trends, confidence calibration, gate effectiveness, and regression rate.
theory-fidelity
Audit theory-to-mechanism mapping fidelity within Mycelium framework projects.
handoff
Generate structured handoff briefs with phone-friendly capture templates.
receipt-render
Generates a one-page work document from completed Diamonds in Claude Code.
mocked-persona-interview
Generate product development personas with speculative evidence tagging.
user-needs-map
Identify, score, and classify user needs for strategic planning.
metrics-detect
Detect and configure external metric sources for product analytics.
Frequently Asked Questions
FAQPage SchemaHow to install Mycelium?βΌ
Run `npx skills add haabe/mycelium --all -g -y` in your terminal to install the full skill set globally.
What problem does Mycelium solve?βΌ
It stops AI coding agents from building products nobody asked for by forcing a short discovery brief and evidence gates before any code is written. Assumptions, confidence scores, and decisions are recorded in your repo so you can audit why anything was built.
How does Mycelium validate a product idea?βΌ
It runs a 10-minute structured interview covering who the product is for, the biggest assumption, and the smallest test, then writes the answers as a versioned brief. Follow-up skills design assumption tests, user interviews, and opportunity trees grounded in real evidence.
Does Mycelium work with agents other than Claude Code?βΌ
Yes. The canvas files and skill instructions are plain markdown and YAML that any agent can read, though automatic hooks and gates currently only fire in Claude Code.
Can I use Mycelium on an existing codebase?βΌ
Yes. The adopt skill reads your existing repo, drafts what the code can establish about delivery and solution shape, and then runs discovery on the gaps it cannot fill, like purpose and real user evidence.