skill

Name: skill
Author: pandazki

152stars15forks

How to install skill?

pandazki/pneuma-skills/skill

$npx -y skills add pandazki/pneuma-skills --skill skill

Installs into the current project.

Summary

AI-orchestrated video production on @pneuma-craft. Use whenever the user wants to generate, edit, or compose video clips, audio tracks, captions, or background music — including text-to-video / image-to-video generation, TTS narration, music generation, provenance tracking, and t

Files · 1

View on GitHub

SKILL.md

1	# ClipCraft
2
3	ClipCraft is a video-production mode where the **source of truth is a
4	structured domain model**, not a file. The in-memory model is an
5	event-sourced craft store from `@pneuma-craft`: an Asset registry, a
6	Composition with Tracks and Clips, and a Provenance DAG that tracks
7	how each asset was generated (and from what). The file `project.json`
8	at the workspace root is a projection of that store — when you edit
9	it with Write/Edit, the viewer auto-re-hydrates. No reload, no
10	refresh signal.
11
12	ClipCraft is built for AIGC workflows: assets are generated, not
13	uploaded. You orchestrate image / video / TTS / BGM generation by
14	running bundled scripts, then record the lineage in `project.json`.
15
16	## Working with the viewer
17
18	The viewer is the exploded 3D timeline rendering of `project.json`.
19	It is the user's source of truth for what's currently selected, what
20	the playhead is on, and what they're pointing at. Four channels link
21	the three actors (you, the user, the viewer):
22
23	### Reading what the user sees
24
25	Every user message arrives wrapped in `<viewer-context>` and (if the
26	user just clicked / dragged / seeked) `<user-actions>`. Read them
27	before you act. Typical clipcraft payloads:
28
29	- `<viewer-context>` — `selectedClipId`, `selectedAssetId`,
30	`selectedTrackId`, `playheadTime` (seconds), `composition.duration`,
31	the active asset's metadata. Use this to disambiguate a vague
32	request like "try another take" — there's almost always a clip
33	selected that tells you which one.
34	- `<preview-frames>` (nested inside `<viewer-context>`) — summary of
35	the planning layer: `total="N"` attribute + per-track `<track id="..."
36	name="..." count="..." />` lines for tracks that carry preview frames.
37	When `total="0"` (or the tag is absent), the timeline has no
38	planning layer yet — a fresh project. Use this to decide whether
39	to start with sketch overlay (see `references/storyboard-workflow.md`)
40	vs jumping straight to generation.
41	- `<user-actions>` — recent UI events: `playhead:seek`
42	(`{time}`), `clip:select` (`{clipId, trackId}`),
43	`asset:select` (`{assetId}`), `clip:drag` (`{clipId, startTime,
44	trackId}`), `track:toggle-mute`, `track:toggle-visible`. Treat
45	these as hints, not commands — the user usually expects you to
46	read them and act, not echo them back.
47
48	If both are absent (cold start, command-button click), inspect
49	`project.json` directly and ask if intent is ambiguous.
50
51	### Locator cards
52
53	After creating or editing assets, clips, or moving the playhead,
54	embed `<viewer-locator>` cards so the user can jump straight to the
55	change. Emit one card per distinct thing you changed — a newly
56	generated asset, a clip you just placed, a time beat you built
57	around — not one per response. The user sees these as clickable
58	chips in chat. Use short concrete labels — "新的 VO 开场",
59	"panda clip on Main", "3.5s — punchline beat" — not generic ones
60	like "see asset".
61
62	#### ViewerAddress — naming an object in the preview
63
64	A `<viewer-locator>` carries an address — a small JSON object that
65	names one referent inside the viewer. clipcraft's address vocabulary
66	is timeline-oriented; every key below is coarse (each one stands
67	alone and drives navigate-then-flash — clipcraft has no element-level
68	fine handle):
69
70	\| Key \| Coarse/fine \| Names \|
71	\|---\|---\|---\|
72	\| `clipId` \| coarse \| a clip on a track — selects it + seeks the playhead to its start \|
73	\| `previewFrameId` \| coarse \| a planning sketch / anchor frame — selects its asset + seeks to its anchor time \|
74	\| `assetId` \| coarse \| an entry in the asset library — scrolls + flashes the tile/row \|
75	\| `trackId` \| coarse \| a track header — scrolls + flashes it \|
76	\| `time` \| coarse \| a moment on the timeline — seeks the playhead only \|
77
78	Supply exactly one key per address. `clipId` / `previewFrameId` /
79	`assetId` / `trackId` are stable ids (survive move / re-pace), so
80	locator references stay valid; `time` is a raw second.
81
82	```html
83	<!-- assetId — scrolls the asset library to the asset and flashes it. -->
84	<viewer-locator address='{"assetId":"asset-vo-tagline"}'>新的 VO 开场</viewer-locator>
85
86	<!-- clipId — selects the clip on the timeline AND seeks the playhead
87	to its startTime, so the us

People who liked this also viewed

6 picks

Type

skill

152stars15forks

View on GitHub

How to install skill?

pandazki/pneuma-skills/skill

$npx -y skills add pandazki/pneuma-skills --skill skill

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# ClipCraft
2
3	ClipCraft is a video-production mode where the **source of truth is a
4	structured domain model**, not a file. The in-memory model is an
5	event-sourced craft store from `@pneuma-craft`: an Asset registry, a
6	Composition with Tracks and Clips, and a Provenance DAG that tracks
7	how each asset was generated (and from what). The file `project.json`
8	at the workspace root is a projection of that store — when you edit
9	it with Write/Edit, the viewer auto-re-hydrates. No reload, no
10	refresh signal.
11
12	ClipCraft is built for AIGC workflows: assets are generated, not
13	uploaded. You orchestrate image / video / TTS / BGM generation by
14	running bundled scripts, then record the lineage in `project.json`.
15
16	## Working with the viewer
17
18	The viewer is the exploded 3D timeline rendering of `project.json`.
19	It is the user's source of truth for what's currently selected, what
20	the playhead is on, and what they're pointing at. Four channels link
21	the three actors (you, the user, the viewer):
22
23	### Reading what the user sees
24
25	Every user message arrives wrapped in `<viewer-context>` and (if the
26	user just clicked / dragged / seeked) `<user-actions>`. Read them
27	before you act. Typical clipcraft payloads:
28
29	- `<viewer-context>` — `selectedClipId`, `selectedAssetId`,
30	`selectedTrackId`, `playheadTime` (seconds), `composition.duration`,
31	the active asset's metadata. Use this to disambiguate a vague
32	request like "try another take" — there's almost always a clip
33	selected that tells you which one.
34	- `<preview-frames>` (nested inside `<viewer-context>`) — summary of
35	the planning layer: `total="N"` attribute + per-track `<track id="..."
36	name="..." count="..." />` lines for tracks that carry preview frames.
37	When `total="0"` (or the tag is absent), the timeline has no
38	planning layer yet — a fresh project. Use this to decide whether
39	to start with sketch overlay (see `references/storyboard-workflow.md`)
40	vs jumping straight to generation.
41	- `<user-actions>` — recent UI events: `playhead:seek`
42	(`{time}`), `clip:select` (`{clipId, trackId}`),
43	`asset:select` (`{assetId}`), `clip:drag` (`{clipId, startTime,
44	trackId}`), `track:toggle-mute`, `track:toggle-visible`. Treat
45	these as hints, not commands — the user usually expects you to
46	read them and act, not echo them back.
47
48	If both are absent (cold start, command-button click), inspect
49	`project.json` directly and ask if intent is ambiguous.
50
51	### Locator cards
52
53	After creating or editing assets, clips, or moving the playhead,
54	embed `<viewer-locator>` cards so the user can jump straight to the
55	change. Emit one card per distinct thing you changed — a newly
56	generated asset, a clip you just placed, a time beat you built
57	around — not one per response. The user sees these as clickable
58	chips in chat. Use short concrete labels — "新的 VO 开场",
59	"panda clip on Main", "3.5s — punchline beat" — not generic ones
60	like "see asset".
61
62	#### ViewerAddress — naming an object in the preview
63
64	A `<viewer-locator>` carries an address — a small JSON object that
65	names one referent inside the viewer. clipcraft's address vocabulary
66	is timeline-oriented; every key below is coarse (each one stands
67	alone and drives navigate-then-flash — clipcraft has no element-level
68	fine handle):
69
70	\| Key \| Coarse/fine \| Names \|
71	\|---\|---\|---\|
72	\| `clipId` \| coarse \| a clip on a track — selects it + seeks the playhead to its start \|
73	\| `previewFrameId` \| coarse \| a planning sketch / anchor frame — selects its asset + seeks to its anchor time \|
74	\| `assetId` \| coarse \| an entry in the asset library — scrolls + flashes the tile/row \|
75	\| `trackId` \| coarse \| a track header — scrolls + flashes it \|
76	\| `time` \| coarse \| a moment on the timeline — seeks the playhead only \|
77
78	Supply exactly one key per address. `clipId` / `previewFrameId` /
79	`assetId` / `trackId` are stable ids (survive move / re-pace), so
80	locator references stay valid; `time` is a raw second.
81
82	```html
83	<!-- assetId — scrolls the asset library to the asset and flashes it. -->
84	<viewer-locator address='{"assetId":"asset-vo-tagline"}'>新的 VO 开场</viewer-locator>
85
86	<!-- clipId — selects the clip on the timeline AND seeks the playhead
87	to its startTime, so the us

People who liked this also viewed

6 picks

Type