extract-source-sample

Name: extract-source-sample
Author: gooseworks-ai

1.0kstars193forks

How to install extract-source-sample?

gooseworks-ai/goose-skills/extract-source-sample

$npx -y skills add gooseworks-ai/goose-skills --skill extract-source-sample

Installs into the current project.

Summary

Given the path to a finished content-goose ad-run folder, extract everything that defines that ad — recipe shot list, VO script, characters, voices, world, atom-skills, master mp4 — and emit a source-sample.json in the exact shape the upload-ad-sample skill writes to the Goos

Files · 1

View on GitHub

SKILL.md

1	# extract-source-sample
2
3	This is an agent-executed skill. There are no Python scripts. The agent
4	reads the run folder, builds the JSON, and stamps catalog links by hand. The
5	content-goose run folders aren't always cleanly structured (some have empty
6	production/ JSON, some carry everything in working/) — an agent adapts, a
7	script would brittle out.
8
9	## When to use
10
11	- "Extract the source-sample.json for `<run>`."
12	- "Get the upload-sample JSON for this ad so I can remix it."
13	- "Prep `<run>` for remix."
14
15	Do NOT use to:
16	- Rewrite the script for a new brand (that's a separate agent step that
17	consumes this skill's output).
18	- Render the remix (that's the existing `remix-ad` skill).
19	- Upload an ad to the library (that's `upload-ad-sample`).
20
21	## Inputs
22
23	\| Input \| Required \| Notes \|
24	\|---\|---\|---\|
25	\| `run-dir` \| yes \| Absolute path to a content-goose ad-run folder (e.g. `clients/ladder/ad-runs/run-02-podcast-skit`). \|
26	\| `out` \| no \| Where to write the JSON. Default: `<run-dir>/remix/source-sample.json`. \|
27
28	That's the entire interface.
29
30	## What the agent must do
31
32	### 1. Read the run
33
34	Open each file if it exists; tolerate missing files (most production/*.json
35	in older runs are empty stubs — fall back to `working/`):
36
37	- `working/script.json` — primary source of truth for scenes, voices, set.
38	- `production/asset-manifest.json` — `assets[]` with role `active_master`
39	points at the master mp4; per-asset `provider` + `metadata.model` produce
40	the atom-skill rows.
41	- `HOW_TO_MAKE_THIS_VIDEO.md` — gets dumped verbatim into `how_to`.
42	- `video-project.json` — fallback for title / format when script.json doesn't
43	carry them.
44	- `finals/*.mp4` — fallback for the master mp4 if asset-manifest is empty.
45	- `working/characters/*.png` — anchor portraits per character.
46	- `working/*.py` — driver scripts (`render_vo.py`, `render_variants.py`,
47	`render_clips.py`, `stitch.py`, `build_end_card.py`, etc.). These are the
48	source's runnable code; the remix consumer ports them. Capture in
49	`production_scripts[]` (step 2 below).
50
51	For sources with character-pose stills (any run with a
52	`working/characters/` folder of `<character>-<pose>.png` files —
53	podcast-skit, founder-led, testimonial, recreate-ugc, etc.), audit every
54	PNG with `file`. Do NOT stop at the base portraits. The recipe shot list
55	references variant expression PNGs (e.g. `brittney-eyebrow-up.png`,
56	`brad-phone-up.png`) by filename; the consumer assumes they exist on disk
57	and will spend real money on lipsync calls before discovering they don't.
58
59	For sources without character-pose stills (music-video b-roll, abstract
60	animated, product-only) — skip this audit; `variant_assets[]` stays empty.
61
62	Record per file (when auditing):
63
64	```jsonc
65	{
66	"file": "brittney-eyebrow-up.png",
67	"pose_tag": "skeptical-eyebrow", // slug from filename stem
68	"kind": "real" \| "lfs-pointer" \| "missing",
69	"size_bytes": 142336
70	}
71	```
72
73	`file <path>` says `PNG image data, …` for real binaries and `ASCII text`
74	for LFS pointers. A real binary is `>10KB` in practice; an LFS pointer is
75	`<200 bytes`.
76
77	Materialize LFS pointers before reading any binary. An LFS pointer is a
78	tiny (<200 byte) ASCII file beginning with `version https://git-lfs.github.com`.
79	If a PNG or mp4 looks like one, run:
80
81	```bash
82	cd <run-dir-or-repo-root>
83	git lfs fetch --include=<relative path>
84	git lfs checkout <relative path>
85	```
86
87	before referencing it. **If `git lfs pull` no-ops and the LFS endpoint
88	returns 404** (objects committed as pointers but never pushed — common on
89	content-goose), leave the entry as `kind: "lfs-pointer"` in
90	`variant_assets[]`. The consumer will regenerate or scrape; this skill
91	does NOT fabricate. See [[feedback_lfs_pointer_audit_before_paid_calls]]
92	and [[feedback_fal_subscribe_error_envelope]] for the downstream cost when
93	this audit is skipped — Hume run-03 lost ~$3 + 25 min to it.
94
95	### 2. Build `source-sample.json`
96
97	Shape (every key always present, arrays may be empty):
98
99	```jsonc
100	{
101	"title": "<from script.json or video-project.json>",
102	"format": "video",
103	"ratio": "<aspect_ratio from script.json — e.g. 9:16>",
104	"formatProfile": "podcast-skit-fabricated", // enum, see below
105	"media_url": "fil

People who liked this also viewed

6 picks

Type

extract-source-sample

1.0kstars193forks

View on GitHub

How to install extract-source-sample?

gooseworks-ai/goose-skills/extract-source-sample

$npx -y skills add gooseworks-ai/goose-skills --skill extract-source-sample

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# extract-source-sample
2
3	This is an agent-executed skill. There are no Python scripts. The agent
4	reads the run folder, builds the JSON, and stamps catalog links by hand. The
5	content-goose run folders aren't always cleanly structured (some have empty
6	production/ JSON, some carry everything in working/) — an agent adapts, a
7	script would brittle out.
8
9	## When to use
10
11	- "Extract the source-sample.json for `<run>`."
12	- "Get the upload-sample JSON for this ad so I can remix it."
13	- "Prep `<run>` for remix."
14
15	Do NOT use to:
16	- Rewrite the script for a new brand (that's a separate agent step that
17	consumes this skill's output).
18	- Render the remix (that's the existing `remix-ad` skill).
19	- Upload an ad to the library (that's `upload-ad-sample`).
20
21	## Inputs
22
23	\| Input \| Required \| Notes \|
24	\|---\|---\|---\|
25	\| `run-dir` \| yes \| Absolute path to a content-goose ad-run folder (e.g. `clients/ladder/ad-runs/run-02-podcast-skit`). \|
26	\| `out` \| no \| Where to write the JSON. Default: `<run-dir>/remix/source-sample.json`. \|
27
28	That's the entire interface.
29
30	## What the agent must do
31
32	### 1. Read the run
33
34	Open each file if it exists; tolerate missing files (most production/*.json
35	in older runs are empty stubs — fall back to `working/`):
36
37	- `working/script.json` — primary source of truth for scenes, voices, set.
38	- `production/asset-manifest.json` — `assets[]` with role `active_master`
39	points at the master mp4; per-asset `provider` + `metadata.model` produce
40	the atom-skill rows.
41	- `HOW_TO_MAKE_THIS_VIDEO.md` — gets dumped verbatim into `how_to`.
42	- `video-project.json` — fallback for title / format when script.json doesn't
43	carry them.
44	- `finals/*.mp4` — fallback for the master mp4 if asset-manifest is empty.
45	- `working/characters/*.png` — anchor portraits per character.
46	- `working/*.py` — driver scripts (`render_vo.py`, `render_variants.py`,
47	`render_clips.py`, `stitch.py`, `build_end_card.py`, etc.). These are the
48	source's runnable code; the remix consumer ports them. Capture in
49	`production_scripts[]` (step 2 below).
50
51	For sources with character-pose stills (any run with a
52	`working/characters/` folder of `<character>-<pose>.png` files —
53	podcast-skit, founder-led, testimonial, recreate-ugc, etc.), audit every
54	PNG with `file`. Do NOT stop at the base portraits. The recipe shot list
55	references variant expression PNGs (e.g. `brittney-eyebrow-up.png`,
56	`brad-phone-up.png`) by filename; the consumer assumes they exist on disk
57	and will spend real money on lipsync calls before discovering they don't.
58
59	For sources without character-pose stills (music-video b-roll, abstract
60	animated, product-only) — skip this audit; `variant_assets[]` stays empty.
61
62	Record per file (when auditing):
63
64	```jsonc
65	{
66	"file": "brittney-eyebrow-up.png",
67	"pose_tag": "skeptical-eyebrow", // slug from filename stem
68	"kind": "real" \| "lfs-pointer" \| "missing",
69	"size_bytes": 142336
70	}
71	```
72
73	`file <path>` says `PNG image data, …` for real binaries and `ASCII text`
74	for LFS pointers. A real binary is `>10KB` in practice; an LFS pointer is
75	`<200 bytes`.
76
77	Materialize LFS pointers before reading any binary. An LFS pointer is a
78	tiny (<200 byte) ASCII file beginning with `version https://git-lfs.github.com`.
79	If a PNG or mp4 looks like one, run:
80
81	```bash
82	cd <run-dir-or-repo-root>
83	git lfs fetch --include=<relative path>
84	git lfs checkout <relative path>
85	```
86
87	before referencing it. **If `git lfs pull` no-ops and the LFS endpoint
88	returns 404** (objects committed as pointers but never pushed — common on
89	content-goose), leave the entry as `kind: "lfs-pointer"` in
90	`variant_assets[]`. The consumer will regenerate or scrape; this skill
91	does NOT fabricate. See [[feedback_lfs_pointer_audit_before_paid_calls]]
92	and [[feedback_fal_subscribe_error_envelope]] for the downstream cost when
93	this audit is skipped — Hume run-03 lost ~$3 + 25 min to it.
94
95	### 2. Build `source-sample.json`
96
97	Shape (every key always present, arrays may be empty):
98
99	```jsonc
100	{
101	"title": "<from script.json or video-project.json>",
102	"format": "video",
103	"ratio": "<aspect_ratio from script.json — e.g. 9:16>",
104	"formatProfile": "podcast-skit-fabricated", // enum, see below
105	"media_url": "fil

People who liked this also viewed

6 picks

Type