a11y-check
Accessibility audit, scoped to the surfaces a product actually has. Detects web / rendered_markdown / terminal / native_app / video_audio / document /…
Render `.claude/canvas/opportunities.yml` as a diagram or structured list. Read-only. Default format Mermaid mindmap. Consults the attribution registry per `${CLAUDE_PLUGIN_ROOT}/engine/render-conventions.md` consent + privacy gate. Second specialist in the render fleet.
$ npx -y skills add haabe/mycelium --skill ost-render --agent claude-codeHow it fires
How this skill gets triggered: by you, by Claude, or both.
/ost-renderContext preview
The summary Claude sees to decide when to auto-load this skill.
Render `.claude/canvas/opportunities.yml` as a diagram or structured list. Read-only. Default format Mermaid mindmap. Consults the attribution registry per `${CLAUDE_PLUGIN_ROOT}/engine/render-conventions.md` consent + privacy gate. Second specialist in the render fleet.
name: ost-render
description: Render `.claude/canvas/opportunities.yml` as a diagram or structured list. Read-only. Default format Mermaid mindmap. Consults the attribution registry per `${CLAUDE_PLUGIN_ROOT}/engine/render-conventions.md` consent + privacy gate. Second specialist in the render fleet.
metadata:
instruction_budget: "60"
framework_dependency: "mycelium"
framework_dependency_note: "Reads .claude/canvas/opportunities.yml + attribution registry (env var $MYCELIUM_ATTRIBUTION_REGISTRY preferred; .claude/memory/attribution-registry.yml fallback). Standalone use will fail without opportunities.yml present."
identifier_exposure: "YES"Read-only render of `.claude/canvas/opportunities.yml` as a diagram or structured list. Second specialist in the render fleet (after `/mycelium:diamond-render`). See `${CLAUDE_PLUGIN_ROOT}/engine/render-conventions.md` for shared conventions.
**Declared**: YES
| Canvas file | Identifier-bearing fields | Frequency | |---|---|---| | `.claude/canvas/opportunities.yml` | `evidence_sources` (URL-or-name strings); `notes` prose | low (~1 in 10 cites a named individual rather than a URL) | | `$MYCELIUM_ATTRIBUTION_REGISTRY` env var (canonical) or `.claude/memory/attribution-registry.yml` (fallback) | `people:` block with `name`+`consent`+`note` per entry | read-only consultation; never rendered |
Per `${CLAUDE_PLUGIN_ROOT}/engine/render-conventions.md#hard-rule-consent--privacy-gate`. Registry lives in roadmap-private memory by design — upstream Mycelium repo is public; registry contents must not ship there.
OST shape is user-need-shaped (abstract opportunities and solutions), so identifier exposure is incidental rather than intrinsic. Identifier classes that can appear: cohort testers, peer practitioners, named external sources. All three go through the consent gate; external sources are typically `public_ok` already, but registry consultation is non-skippable to avoid drift.
Per `engine/render-conventions.md#anon-label-convention`:
Numbering resets per render; emit an in-render mapping footnote when redaction occurred so the operator can audit which anon label maps to which registry entry.
Per `engine/render-conventions.md#consent-value-semantics`:
| Value | Render behavior | |---|---| | `public_ok` | Render literal. Append carve-out footnote pointer if entry has non-empty `note:`. | | `generic_only` | Redact to anon-label. Queue anon-mapping footnote. | | `unknown` | Treat as `generic_only` per registry README. | | (not in registry) | Fail loud unless `--no-identifiers=true`. |
**public_ok → literal**: `evidence_sources: ["Drew Hoskins LinkedIn 2026-05-11"]`, registry entry `{name: "Drew", consent: public_ok, note: "..."}` → renders as `Drew Hoskins LinkedIn 2026-05-11` + carve-out footnote pointer.
**generic_only → anon-label**: `evidence_sources: ["Daniel Bentes install report"]`, registry entry `{name: "Daniel", consent: generic_only, ...}` → renders as `peer-practitioner-1 install report` + anon-mapping footnote.
**Already-anon canvas label → preserve**: `evidence_sources: ["cohort-tester-3 session 2 transcript"]`, canvas already anon-labeled → renders as `cohort-tester-3 session 2 transcript` (preserve canvas-side anonymization; do not look up).
**Identifier absent from registry → fail loud**: `evidence_sources: ["Random Name DM 2026-06-05"]`, no entry → **the WHOLE render is blocked — emit NO artifact** (not a partial mindmap with the offending entry redacted or omitted), and give the fail-loud message with three fix options (add to registry, re-run with `--no-identifiers`, edit canvas). Clarified v0.56.0 after a dogfood run read this line as per-entry and rendered around the unregistered source: an unresolved consent state means the render's exposure declaration cannot be made honestly, so nothing ships until the user resolves it. Naming the offending source in the owner-facing explanation is fine (project-private context); the artifact is what must not exist.
1. Read `.claude/canvas/opportunities.yml` with the Read tool. Full read for emit; not `limit:1`. 2. Read the attribution registry per path-resolution order in `engine/render-conventions.md#registry-path-resolution`: `$MYCELIUM_ATTRIBUTION_REGISTRY` env var first; fall back to `.claude/memory/attribution-registry.yml`. Registry root key is `people:`; each entry has `name`, `consent`, optional `note`. If registry absent, surface `⚠ no attribution-registry — consent-redaction not enforceable; treat output as roadmap-internal` warning in the render header. **For `--audience external`, an absent registry BLOCKS the render and emits no artifact** (v0.173.0) — same outcome as an unregistered identifier, because an absent registry is that same unresolved-consent condition applied to every identifier at once. The warning-and-emit path is for `founder`/`cohort` only. 3. Note the source's canvas-state timestamp
A harness that asks who this is for before the agent writes code. Built on Claude Code, where the gates are structural. The files and skills port to opencode, Codex and Cursor. Outcome over output. You know how this goes.
Accessibility audit, scoped to the surfaces a product actually has. Detects web / rendered_markdown / terminal / native_app / video_audio / document /…
Bring Mycelium into a project that already has code. Detects that the repo predates the framework, asks before touching anything, then reads the codebase to…
Design the smallest viable test to validate or invalidate a critical assumption. Based on Torres's assumption testing framework, organized by Gilad's AFTER…
Use before any research activity or significant decision. Reviews cognitive biases relevant to the current stage.
Use to evaluate whether current work aligns with Better Value Sooner Safer Happier. Run at diamond completion and periodically.
Lint canvas files for staleness, missing fields, inconsistent evidence types, and orphaned references. Run periodically or before major transitions.