impl-screen-specs
Generate the SCREEN_SPECS manifest — a per-screen specification (components, data, API calls, user actions, states, i18n). Optional tool a developer may use anytime to blueprint screens; not required before code (TASK-230, 2026-09-16 human decision). Vendored Spine skill; see src/manifest-policy.js.
- License 许可
- MIT
- 最近更新
- 2026-10-10
- Stars
- ★ 1
- Forks
- 0
01 前言 Front Matter
- name
- impl-screen-specs
- description
- Generate the SCREEN_SPECS manifest — a per-screen specification (components, data, API calls, user actions, states, i18n). Optional tool a developer may use anytime to blueprint screens; not required before code (TASK-230, 2026-09-16 human decision). Vendored Spine skill; see src/manifest-policy.js.
03 SKILL.md 正文
Sources & manifest status (hivemind)
Vendored from implementation-engine. In hivemind the "context" this skill reads is the project's
knowledge surface — PROJECT.md, the wisearcher brain graph (kb_search/kb_answer, or the local
knowledge/ KB when the brain is offline), and the ticket's scope + acceptance criteria. Where the
body below refers to context/<file>.md, read the equivalent from that surface. Output goes to
implementation/SCREEN_SPECS.md.
Status (TASK-230, 2026-09-16 human decision): optional tool, not a gate. No verification_tier
requires this manifest before code — see src/manifest-policy.js (requiresManifest always returns
false). Generate/update it whenever it helps, in whatever order the Developer chooses; the
reviewer does NOT treat a missing manifest as a finding. Preserve epistemic markers
([EXPLICIT]/[INFERRED:strong|weak]/[ASSUMED]) and respect source tiers.
impl-screen-specs
Generates implementation/SCREEN_SPECS.md — a per-screen specification covering every screen in the project scope.
Read first
Read these files from context/ (and only from context/):
context/scope.md— identifies all scope items (S-##) and their feature areascontext/technical.md— navigation structure, architecture patterns, component conventionscontext/execution.md— phase assignments per feature area
Output: implementation/SCREEN_SPECS.md
Write one entry per screen. Derive the screen list from the scope items (S-AUTH, S-HOME, S-CAMP, S-WORK, S-CHAT, S-PROF, S-SET, S-PAY, S-REP) and expand each into its constituent screens.
Format for each screen
## <Screen Name>
**Scope item**: S-##
**Phase**: P#
**Navigation path**: <tab> → <stack position> (e.g., Campaigns tab → Campaign Detail)
### Components
- <Canonical component name from COMPONENT_CATALOG>: <brief purpose>
(reference catalog entries by name — never describe a bespoke date/select/etc.)
### Data
- <what data this screen needs, where it comes from (API / state store / prop)>
### API calls
- <METHOD /path> — <purpose> [gap: G-## if blocked]
### User actions
- <action> → <result or navigation target> — span `<feature.action>`
### States
- **Loading**: <description>
- **Empty**: <description>
- **Error**: <description> — traced + logged (see Observability standard)
- **Populated**: <description>
### i18n keys
- `<key>`: "<default text>"
### Open gaps
- <G-## or [MISSING_INFO] if something cannot be determined from context>
Rules
- Apply
.claude/shared/UI_CONSISTENCY.md: reference canonical components by catalog name; flag any screen that would need a bespoke variant of a shared input. - Apply
.claude/shared/OBSERVABILITY.md: every user action names its trace span; the Error state is traced + logged. - Cover every screen derivable from scope. If a scope item maps to multiple screens (e.g., S-WORK has 7 workflow steps), create one entry per step.
- Use generic terms for architecture references ("state store", "HTTP client") except when an ADR decision is being cited — in that case label it clearly.
- Do not invent API endpoints. If the endpoint is unknown, write
[MISSING_INFO] endpoint TBDand cite the relevant gap. - Mark any uncertainty with
[INFERRED]or[ASSUMED]as appropriate. - Preserve
[MISSING_INFO]for anything that cannot be derived from context.
After writing
Report: total screen count, scope items covered, number of [MISSING_INFO] markers found.
04 同仓库更多 Skills
unknown-skill
A demo skill with no license signal at all, used to exercise the "unknown" classification branch of the human license gate (TASK-135).
spoofed-license-skill
A demo skill used to test TASK-150's SPDX-header-vs-LICENSE-file conflict surfacing.
permissive-skill
A demo skill used to test assimilation of a permissively-licensed third party skill.
malicious-skill
A fixture skill deliberately carrying risky content patterns, used only to exercise src/skill-scan.js (TASK-122). Never assimilated for real.
copyleft-skill
A demo skill used to test the human license gate for a copyleft third party skill.
watch
Watch a video (URL or local path). Downloads with yt-dlp, extracts auto-scaled frames with ffmpeg, pulls the transcript from captions (or Whisper API fallback), and hands the result to Claude so it can answer questions about what's in the video. (assimilated for Hivemind)