cycle-render
Render `.claude/canvas/cycle-history.yml` as a Gantt chart + outcome distribution pie (or ascii / json). Read-only. Consults the attribution registry per `${CLAUDE_PLUGIN_ROOT}/engine/render-conventions.md` consent + privacy gate. Third specialist in the render fleet.
haabe · 已有静态检查记录
安装与使用
运行命令前先阅读来源与依赖说明。复制命令不会自动安装或执行。
npx skills add haabe/mycelium --skill cycle-render技能说明
name: cycle-render
description: Render .claude/canvas/cycle-history.yml as a Gantt chart + outcome distribution pie (or ascii / json). Read-only. Consults the attribution registry per ${CLAUDE_PLUGIN_ROOT}/engine/render-conventions.md consent + privacy gate. Third specialist in the render fleet.
metadata:
instruction_budget: "55"
framework_dependency: "mycelium"
framework_dependency_note: "Reads .claude/canvas/cycle-history.yml + attribution registry (env var $MYCELIUM_ATTRIBUTION_REGISTRY preferred; .claude/memory/attribution-registry.yml fallback)."
identifier_exposure: "YES"
Cycle Render
Read-only render of .claude/canvas/cycle-history.yml as gantt + pie + (optionally) json. Third specialist in the render fleet. See ${CLAUDE_PLUGIN_ROOT}/engine/render-conventions.md for shared conventions.
When NOT to use
- For cross-cutting opportunity→solution→cycle traceability view → dispatcher's
/mycelium:render --view traceability(deferred to Phase 4a–4d research-first methodology). - To RECORD a new cycle outcome →
/mycelium:retrospective,/mycelium:ice-score(on discard), or/mycelium:launch-tier(on launch). This skill is read-only. - For framework-health full assessment →
/mycelium:framework-health. cycle-render emits one visualization piece; framework-health is the broader audit.
Identifier exposure
Declared: YES
Scope (canvas surfaces touched)
| Canvas file | Identifier-bearing fields | Frequency |
|---|---|---|
.claude/canvas/cycle-history.yml | learnings.process prose; learnings.framework prose; related_corrections references | mid (cohort-related observation cycles directly reference participants) |
$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 |
Rationale
Cycle history is reflection-shaped: learnings.process is prose-heavy and often names testers whose feedback drove the cycle (e.g., a cycle entry with leaf_id: ht-014-alex-first-week-observation carries an identifier in the leaf_id itself). Identifier exposure is structural to the learnings narrative, not incidental. Consent state for named cohort testers can shift over time per the consent-state-change-skip cluster (anti-pattern #7 sub-shape #15) — cycle-render re-consults the registry on every invocation, never caches.
Anon-label convention
Per engine/render-conventions.md#anon-label-convention. Numbering shared across the render-fleet session (cohort-tester-N in cycle-003 IS cohort-tester-N in opp-004 evidence if both rendered in one session).
In gantt output, anon labels appear in section names and task IDs. In ascii output, in chronological listings. In json output, in learnings_process strings AND a parallel identifier_map field for auditability.
Consent value semantics
Per engine/render-conventions.md#consent-value-semantics. public_ok → render literal + carve-out footnote pointer if entry has non-empty note:. generic_only → redact to anon-label. unknown → treat as generic_only. Not-in-registry → fail loud unless --no-identifiers=true.
Worked examples
public_ok cohort tester → literal: cycle entry with leaf_id ht-014-alex-first-week-observation, registry entry {name: "Alex", consent: public_ok, note: "..."} → gantt section reads Alex first-week observation + carve-out footnote pointer.
generic_only → anon-label: a future cycle naming "Random Tester" with registry consent: generic_only → gantt section reads cohort-tester-N first-week observation + anon-mapping footnote.
Maintainer literal: a cycle prose mentions "Håvard's Torres-shape question", registry {name: "Håvard", consent: public_ok} → renders literally.
Identifier not in registry → fail loud: a future cycle adds a tester never registered → fail-loud per engine/render-conventions.md. The upstream fix is in /mycelium:retrospective (consult registry on cycle recording); cycle-render is the downstream gate.
Fixture pointer
tests/bash/fixtures/cycle-render/redaction-public-ok-literal-gantt.ymltests/bash/fixtures/cycle-render/redaction-generic-only-anon-gantt.ymltests/bash/fixtures/cycle-render/redaction-public-maintainer.ymltests/bash/fixtures/cycle-render/redaction-no-registry-entry-fail-loud.ymltests/bash/fixtures/cycle-render/redaction-carve-out-note-footnote.yml
Preflight: Read sources
- Read
.claude/canvas/cycle-history.ymlwith the Read tool. Full read; notlimit:1. - Read the attribution registry per path resolution order in
engine/render-conventions.md#registry-path-resolution:$MYCELIUM_ATTRIBUTION_REGISTRYenv var first; fall back to.claude/memory/attribution-registry.yml. If registry absent, surface a⚠ no attribution-registry — consent-redaction not enforceable; treat output as roadmap-internalwarning 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 forfounder/cohortonly. - Note canvas-state timestamp per
engine/render-conventions.md#canvas-state-timestamp-resolution:_meta.last_validatedif present; fall back to top-levellast_updated:.
Arguments
| Arg | Default | Values | Effect |
|---|---|---|---|
--format | mermaid | mermaid | ascii | json | Output format. markdown-table and markdown-list are NOT supported (gantt doesn't map cleanly); fail loud per engine/render-conventions.md#format-support-negotiation-global-rule. |
--view | both | gantt | pie | both | Which diagram(s) to emit. |
--theme | base | base | dark | Theme. dark is the WCAG-by-construction opt-in per `engine/render-conventi |