# Skills index

**Audience**: practitioners looking up a skill, evaluators surveying capability surface.
**Time to read**: 5 min for the table; depth lives in each `SKILL.md`.
**Last updated**: 2026-06-12.

This index lists all 63 skills. Each skill's full behaviour lives in its `SKILL.md` — `plugins/mycelium/skills/<name>/SKILL.md` in plugin form (recommended), or `.claude/skills/<name>/SKILL.md` in legacy form (supported during transition). Loaded JIT by Claude Code. This page is the one skills index, grouped by the scale each skill mostly serves.

## Onboarding & navigation

| Skill | When to use | Gates it satisfies |
|---|---|---|
| `/interview` | New project: purpose, vision, North Star, project classification | Evidence (L0) |
| `/diamond-assess` | Resume a session: where am I and what next | — |
| `/diamond-progress` | Move a diamond forward through gate checks | All applicable to the transition |
| `/define-done` | Pin a diamond's outcome Definition of Done (behaviour-change, not a build-list) at birth or retrofit | Outcomes (DoD) |

## Discovery (L0–L2)

| Skill | When to use | Gates it satisfies |
|---|---|---|
| `/user-interview` | Story-based interviews with bias mitigation (Torres) | Evidence, JTBD |
| `/mocked-persona-interview` | Disciplined mocked personas when real users unavailable | Evidence (with speculation tag) |
| `/user-needs-map` | Map needs independently of solutions (Allen) | Evidence |
| `/jtbd-map` | Jobs to be Done mapping — functional, emotional and social dimensions (Christensen) | JTBD |
| `/cynefin-classify` | Classify problem domain (Snowden) | Domain Fit |
| `/wardley-map` | Strategic landscape mapping (L1) | Evidence (L1) |
| `/ost-builder` | Build Opportunity Solution Tree from research (L2) | Evidence, Four Risks (per leaf) |
| `/assumption-test` | Design smallest viable test for an assumption (Torres + Gilad AFTER) | Four Risks |
| `/ice-score` | Prioritization by Impact x Confidence x Ease, with Gilad's Confidence Meter (the average, Ellis's arithmetic, until v0.318.x) | — |
| `/handoff` | Structured handoff for offline human tasks (interviews, observations, outreach) | — |
| `/log-evidence` | Record findings from completed offline conversations back into canvas | Evidence |

## Solution & planning (L3)

| Skill | When to use | Gates it satisfies |
|---|---|---|
| `/gist-plan` | GIST planning: goals, ideas, steps, tasks (Gilad) | — |
| `/preflight` | Pre-implementation validation checklist | Learning |

## Build & delivery (L4)

| Skill | When to use | Gates it satisfies |
|---|---|---|
| `/delivery-bootstrap` | Auto-detect tech stack, generate validation tooling, scaffold ADRs | — |
| `/reflexion` | Self-correcting loop: implement → validate → self-critique → retry (max 3) | — |
| `/definition-of-done` | Executable checklist: tests, types, lint, secrets, BVSSH, accessibility, decision log | Outcomes (BVSSH), Service Quality |
| `/dora-check` | DORA metrics (deployment freq, lead time, CFR, FDRT, reliability) | Delivery Health |
| `/retrospective` | Post-delivery: what worked, what did not, pattern + correction extraction | Learning |

## Audit & governance (any scale)

| Skill | When to use | Gates it satisfies |
|---|---|---|
| `/bias-check` | Review cognitive biases before research/decisions | Bias |
| `/devils-advocate` | Challenge assumptions before major decisions (counter-argument) | Bias |
| `/bvssh-check` | Holistic BVSSH outcome evaluation | Outcomes (BVSSH) |
| `/service-check` | Downe's 15 service principles | Service Quality |
| `/threat-model` | STRIDE threat modeling | Security |
| `/privacy-check` | Privacy by Design / GDPR (Cavoukian) | Privacy |
| `/security-review` | OWASP Top 10:2025 secure-design review | Security |
| `/harms-check` | Design-level harm to users: dark/deceptive patterns + foreseeable misuse (safety-by-design; NUDGE). Named by the release path since 0.321.0, as advice: its verdict is recorded on the exposure record and never blocks | Harms (BVSSH Happier) |
| `/usability-check` | Nielsen's 10 usability heuristics | Service Quality |
| `/a11y-check` | Accessibility audit (WCAG 2.1 AA) | Service Quality |
| `/regulatory-review` | EU AI Act risk classification (Annex III, Art 13/50) | Regulatory |
| `/xai-check` | Explainability audit for AI-containing products (5-stage tier-scaled) | Explainability |
| `/wiring-contract` | Check that declared mechanisms are actually reached — producers have consumers, and readers resolve the paths writers write | — |

## Evidence & metrics

| Skill | When to use | Gates it satisfies |
|---|---|---|
| `/metrics-detect` | Detect external metric sources (GitHub, analytics, payments, app stores) | — |
| `/metrics-pull` | Snapshot all configured sources, compute deltas, draft canvas evidence | Evidence |

## Market & organization (L5 / L1)

| Skill | When to use | Gates it satisfies |
|---|---|---|
| `/launch-tier` | Classify releases, plan go-to-market (Lauchengco) | — |
| `/team-shape` | Team Topologies assessment (Skelton) | — |

## Canvas & orchestration

| Skill | When to use | Gates it satisfies |
|---|---|---|
| `/canvas-update` | Update canvas with new evidence | Evidence |
| `/canvas-health` | Lint canvas for staleness, missing fields, orphaned references | — |
| `/canvas-sync` | Synchronize canvas across team via git (cross-session sync helper) | — |
| `/fan-out` | Parallel agent orchestration with worktree isolation | — |

## Render & output (v0.40.0+)

Read-only rendering of canvas + state surfaces. All five skills share `engine/render-conventions.md` (consent + privacy HARD RULE, supported formats, WCAG AA theme convention, frontmatter Mermaid syntax) and Validator Check 43's `identifier_exposure: YES|NONE|MIXED` declaration. The first three plus the dispatcher are the canvas-visualization fleet; `/receipt-render` shares the conventions but is a standalone shareable-artifact generator, not a `/render` target.

| Skill | When to use | Gates it satisfies |
|---|---|---|
| `/diamond-render` | Emit `diamonds/active.yml` as Mermaid stateDiagram-v2 / ascii / json (the decisions each diamond has recorded, grouped by mode). Recommend at end of `/diamond-assess`. | — |
| `/ost-render` | Emit `opportunities.yml` as Mermaid mindmap / ascii / markdown-list / json. Consent-gate via attribution registry. | — |
| `/cycle-render` | Emit `cycle-history.yml` as Mermaid gantt + pie / ascii / json. Honest small-N + class-distribution disclosure. | — |
| `/render` | Dispatcher: routes intent to a specialist (recommends, never auto-invokes); lists what's renderable. Cross-cutting `--view traceability` research-gated to Phase 4a–4d. | — |
| `/receipt-render` | Turn a completed diamond into a shareable, factual one-page work receipt (problem framed, assumptions tested, killed vs kept) with a volitional, no-telemetry onward-handoff data flow. Standalone; not a `/render` target. | — |

## Setup & lifecycle

| Skill | When to use | Gates it satisfies |
|---|---|---|
| `/start` | First command after installing the plugin: composes `/setup` + `/interview` into one flow — the recommended entry point | — |
| `/setup` | Initialize project-state directories + starter files only (idempotent); use standalone when you want state without the interview | — |
| `/migrate-from-legacy` | Move a legacy (npx-degit) install to plugin form; idempotent, verifies project state survived | — |
| `/adopt` | Two-phase on-ramp for an existing codebase: populates what the code can establish, then patches the holes it cannot with the same discipline as greenfield. Since 0.322.0 it asks whether something is already in front of real people when the repository shows the project deploys, and drafts the record of who it reaches | — |
| `/ping` | Smoke-test that the plugin loaded correctly (deterministic marker); not for normal end-user use | — |

## Self-improvement (framework-level)

| Skill | When to use | Gates it satisfies |
|---|---|---|
| `/feedback-review` | Aggregate feedback signals across active loops, check health | — |
| `/eval-runner` | Run benchmark scenarios to measure framework effectiveness | — |
| `/corrections-audit` | Analyze correction trends, surface recurring patterns, flag graduation candidates | — |
| `/friction-log` | Draft an account of where *using* Mycelium was confusing, for the user to edit and choose whether to send (Hoskins) | — |
| `/prompt-optimizer` | A/B test instruction changes against eval benchmarks | — |
| `/framework-health` | Quarterly self-assessment: cycle velocity, discard trends, calibration, regression rate | — |
| `/theory-fidelity` | Audit whether the theories the project claims are faithfully operationalized (source-grounds load-bearing ones); pair with `/framework-health` quarterly | — |
| `/scaffold-cost-check` | Measure Mycelium's own scaffold token cost (CLAUDE.md + engine + harness + canvas + memory); pair with `/framework-health` for trend | — |

## How to use this index

If you are mid-diamond and want a skill: look at the section that matches your scale. If you are auditing: look at audit & governance. If you are doing framework-level work: look at self-improvement.

Skills are auto-discovered from `.claude/skills/*/SKILL.md` frontmatter — Claude Code reads them at session start. Adding a skill is creating a new directory + SKILL.md; this index then needs updating per the [version-discipline](../../plugins/mycelium/engine/version-discipline.md) rule on material framework changes.

## See also

- [glossary.md](../glossary.md) — vocabulary used in skill descriptions
- `plugins/mycelium/engine/theory-gates.md` — canonical gate definitions