@cliangdev/conductor 0.41.1 → 0.42.1
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/assets/claude/agents/conductor-coder.md +278 -7
- package/assets/claude/agents/{conductor-content-art-director.md → conductor-marketing-art-director.md} +2 -2
- package/assets/claude/agents/{conductor-content-creative-director.md → conductor-marketing-creative-director.md} +12 -8
- package/assets/claude/agents/{conductor-content-ideator.md → conductor-marketing-ideator.md} +11 -5
- package/assets/claude/agents/{conductor-content-scriptwriter.md → conductor-marketing-scriptwriter.md} +19 -6
- package/assets/claude/agents/{conductor-content-strategist.md → conductor-marketing-strategist.md} +2 -2
- package/assets/claude/commands/conductor/implement.md +1 -1
- package/assets/claude/commands/conductor/marketing.md +26 -0
- package/assets/claude/skills/conductor-marketing-create/SKILL.md +87 -0
- package/assets/claude/skills/{conductor-creative → conductor-marketing-create}/references/brand-kit-setup.md +2 -3
- package/assets/claude/skills/{conductor-content-studio → conductor-marketing-create}/references/design-critique.md +0 -1
- package/assets/claude/skills/{conductor-creative/SKILL.md → conductor-marketing-create/references/direct.md} +50 -60
- package/assets/claude/skills/{conductor-creative → conductor-marketing-create}/references/hooks.md +13 -9
- package/assets/claude/skills/{conductor-content-studio → conductor-marketing-create}/references/idea-template.md +2 -2
- package/assets/claude/skills/{conductor-creative → conductor-marketing-create}/references/layout-guide.md +0 -1
- package/assets/claude/skills/{conductor-creative → conductor-marketing-create}/references/realism-check.md +0 -1
- package/assets/claude/skills/{conductor-content-studio → conductor-marketing-create}/references/rubric.md +15 -6
- package/assets/claude/skills/{conductor-content-studio → conductor-marketing-create}/references/script-template.md +4 -2
- package/assets/claude/skills/{conductor-creative → conductor-marketing-create}/references/slideshow-formats.md +12 -9
- package/assets/claude/skills/{conductor-content-studio/SKILL.md → conductor-marketing-create/references/team-run.md} +33 -37
- package/assets/claude/skills/conductor-marketing-create/references/voice.md +120 -0
- package/assets/claude/skills/{conductor-publisher → conductor-marketing-publish}/SKILL.md +5 -5
- package/assets/claude/skills/{conductor-marketing-analyst → conductor-marketing-report}/SKILL.md +11 -4
- package/assets/cli-manifest.json +17 -37
- package/dist/lib/plugin-assets.d.ts.map +1 -1
- package/dist/lib/plugin-assets.js +98 -42
- package/dist/lib/plugin-assets.js.map +1 -1
- package/package.json +1 -1
- package/assets/claude/commands/conductor/content.md +0 -66
- package/assets/claude/commands/conductor/creative.md +0 -108
- package/assets/claude/commands/conductor/marketing-report.md +0 -38
- package/assets/claude/skills/conductor-coder/SKILL.md +0 -282
- package/assets/claude/skills/conductor-content-studio/references/layout-guide.md +0 -73
- package/assets/claude/skills/conductor-content-studio/references/realism-check.md +0 -95
- package/assets/claude/skills/conductor-content-studio/references/slideshow-formats.md +0 -135
- package/assets/claude/skills/conductor-creative/references/design-critique.md +0 -52
- package/assets/claude/skills/conductor-ux-ui-design/SKILL.md +0 -346
- package/assets/claude/skills/conductor-ux-ui-design/references/design-tokens.md +0 -359
- /package/assets/claude/skills/{conductor-content-studio → conductor-marketing-create}/references/capabilities.md +0 -0
- /package/assets/claude/skills/{conductor-content-studio → conductor-marketing-create}/references/compliance-checklist.md +0 -0
- /package/assets/claude/skills/{conductor-content-studio → conductor-marketing-create}/references/platform-specs.md +0 -0
- /package/assets/claude/skills/{conductor-content-studio → conductor-marketing-create}/references/publish-pack-template.md +0 -0
- /package/assets/claude/skills/{conductor-content-studio → conductor-marketing-create}/references/shotlist-template.md +0 -0
- /package/assets/claude/skills/{conductor-marketing-analyst → conductor-marketing-report}/references/html-layout.md +0 -0
|
@@ -1,29 +1,34 @@
|
|
|
1
|
-
|
|
2
|
-
|
|
3
|
-
|
|
4
|
-
|
|
5
|
-
|
|
6
|
-
|
|
7
|
-
|
|
8
|
-
|
|
9
|
-
|
|
10
|
-
|
|
11
|
-
|
|
12
|
-
|
|
13
|
-
|
|
14
|
-
|
|
15
|
-
|
|
16
|
-
|
|
17
|
-
|
|
18
|
-
|
|
19
|
-
|
|
20
|
-
|
|
21
|
-
|
|
22
|
-
|
|
23
|
-
|
|
24
|
-
|
|
25
|
-
|
|
26
|
-
|
|
1
|
+
# Direct mode: making a Creative from what is in hand
|
|
2
|
+
|
|
3
|
+
Read when the person already has the headline, photo, clip or finished images (or wants a lettered
|
|
4
|
+
headline variant). The shared rules in `SKILL.md` (brand kit first, draft first, no Post) apply.
|
|
5
|
+
|
|
6
|
+
The copy and the visual language both come from things that already exist for this workspace: a
|
|
7
|
+
Brand Kit (data: tokens, copy rules, approved lines) and a Knowledge page (prose: positioning, voice,
|
|
8
|
+
facts). Read them rather than inventing. Rendering happens locally through this MCP server with a
|
|
9
|
+
browser on this machine; a preview typically finishes in well under a minute.
|
|
10
|
+
|
|
11
|
+
## What to make
|
|
12
|
+
|
|
13
|
+
If the request does not already say, ask once with `AskUserQuestion` (single select), then gather
|
|
14
|
+
only what the choice needs: one or two follow-up questions, not a form.
|
|
15
|
+
|
|
16
|
+
- **Static ad** (photo + headline in the brand layout, at every enabled placement; not for organic
|
|
17
|
+
posts). Needs a headline (or draft one from the Knowledge page) and a photo: call
|
|
18
|
+
`list_creative_media({kind: "IMAGE"})` first and offer what is on hand before asking for a new one.
|
|
19
|
+
Nothing fits and none is offered? Offer `generate_creative_image` (step 5). A new photo is a local
|
|
20
|
+
file path (a URL: download it first); it is only uploaded after the person approves the preview.
|
|
21
|
+
- **Short video from a photo** (animated text over a still or a moving background): same inputs as a
|
|
22
|
+
static ad; see "Short videos from a photo or clip".
|
|
23
|
+
- **Video from a clip I have** (brand layout animated over an existing video) and **post an existing
|
|
24
|
+
video as-is** (no layout): call `list_creative_media({kind: "VIDEO"})` first; if nothing fits, ask
|
|
25
|
+
for a local path (a URL: download it first). A clip used as a MOTION background is previewed from
|
|
26
|
+
the local file too.
|
|
27
|
+
- **Organic post or slideshow** (a TikTok or Instagram photo post that is not an ad): skip the brand
|
|
28
|
+
layouts and follow the slideshow path below. Ask for the topic, the photos and a real app
|
|
29
|
+
screenshot if there is one.
|
|
30
|
+
- **Headline variant** of an existing Creative: ask which one (id or display id like `12a`) and the
|
|
31
|
+
new headline; flow in step 6.
|
|
27
32
|
|
|
28
33
|
## Which path
|
|
29
34
|
|
|
@@ -58,8 +63,8 @@ default, not this workspace's actual brand. Set it up before writing any ad rath
|
|
|
58
63
|
against placeholder tokens: gather colours, font, CTA claim and any copy rules or approved lines
|
|
59
64
|
(from a website via `WebFetch`, a brand doc via `Read`, or by asking directly), then call
|
|
60
65
|
`update_brand_kit` for the data fields and `upload_brand_image` for the mark/wordmark/badge. Confirm
|
|
61
|
-
with `get_brand_kit` afterward — `configured` should now read `true`. (
|
|
62
|
-
a person through exactly this
|
|
66
|
+
with `get_brand_kit` afterward — `configured` should now read `true`. (`references/brand-kit-setup.md`
|
|
67
|
+
walks a person through exactly this.)
|
|
63
68
|
|
|
64
69
|
### 2. Pick the angle
|
|
65
70
|
|
|
@@ -74,7 +79,8 @@ of the problem or the result) before you write a line of copy.
|
|
|
74
79
|
|
|
75
80
|
Prefer an approved line from the Knowledge page verbatim. Otherwise write one that names something
|
|
76
81
|
real and specific, not a generic claim. Write 6 to 10 candidates across hook types, score them and
|
|
77
|
-
keep the best two or three: `references/hooks.md`.
|
|
82
|
+
keep the best two or three: `references/hooks.md`. Every candidate must read like a person posted
|
|
83
|
+
it: `references/voice.md`.
|
|
78
84
|
|
|
79
85
|
If the Brand Kit's `accentPhraseRequired` is true, mark exactly one phrase per headline with
|
|
80
86
|
asterisks: `Every saved link, *finally usable*.` Never zero, never two — `preview_creative_draft`,
|
|
@@ -143,7 +149,7 @@ concealment: the AI label still applies.
|
|
|
143
149
|
**Open the photo and look at it before using it** — a generated one most of all (pseudo-text, warped
|
|
144
150
|
hands or faces, a fake product or app screen; the realism check covers the rest). A picture can break every copy rule with text,
|
|
145
151
|
a logo or a claim burned into its pixels, and no automated check can see that — this is the first
|
|
146
|
-
of this skill's silent failure modes (the full list is
|
|
152
|
+
of this skill's silent failure modes (the full list is in `SKILL.md`). A human or a prior pass marks a bad
|
|
147
153
|
photo `blocked`; if you notice one that should be, say so rather than using it anyway.
|
|
148
154
|
|
|
149
155
|
Match the crop to the subject. A layout that lays copy over the photo (`bleed`) can put a face in
|
|
@@ -252,8 +258,19 @@ Creative is already saved — call `render_creative`, don't commit again. Verify
|
|
|
252
258
|
Then `attach_creative_to_post` with the Post's Work Item id — it maps each placement to the right
|
|
253
259
|
destinations (9:16 to TikTok/Reels, 4:5 to Instagram feed, 1:1 to Facebook, story to story targets)
|
|
254
260
|
for every destination that has not set its own custom media, and never touches one that has. Verify
|
|
255
|
-
with `get_post_status`. From here, use the `conductor-
|
|
256
|
-
publishing — this skill's job ends at attaching the artwork.
|
|
261
|
+
with `get_post_status`. From here, use the `conductor-marketing-publish` skill for scheduling, review and
|
|
262
|
+
publishing — this skill's job ends at attaching the artwork. If there is no Post yet, say so plainly
|
|
263
|
+
and name the next step (creating one with `conductor-marketing-publish`) rather than leaving the artwork with
|
|
264
|
+
no stated destination.
|
|
265
|
+
|
|
266
|
+
When the Creative came from a team run, or the person points to a publish pack, look for
|
|
267
|
+
`.conductor/content-runs/*/publish-pack.json` whose `creativeId` matches. It carries per-destination
|
|
268
|
+
captions and publish options (disclosure, AI label, cover frame, alt text) and a pinned comment. When
|
|
269
|
+
you go on to create or edit the Post with `conductor-marketing-publish`, pre-fill each target's
|
|
270
|
+
`captionOverride` and options from it, show them in the confirmation table and let the person change
|
|
271
|
+
them. Mention the pinned comment (and, for a post that promotes an app, the Stories repost with a link
|
|
272
|
+
sticker) as manual steps, since Conductor cannot post comments. The Post still goes through the normal
|
|
273
|
+
review.
|
|
257
274
|
|
|
258
275
|
## Making a sequence instead of a single ad
|
|
259
276
|
|
|
@@ -347,7 +364,7 @@ thing).
|
|
|
347
364
|
accepts a chosen track on photos, so each destination sends it as `carouselAs: "photos"`
|
|
348
365
|
(silent) or `"video"` (the slideshow MP4 with the track). Attach defaults to video for TikTok
|
|
349
366
|
and YouTube and photos for Instagram and Facebook feed, and writes the choice into the target's
|
|
350
|
-
options; change it with `set_publish_targets` (
|
|
367
|
+
options; change it with `set_publish_targets` (`conductor-marketing-publish` skill). Mind each platform's length limit
|
|
351
368
|
for the video.
|
|
352
369
|
|
|
353
370
|
## Short videos from a photo or clip
|
|
@@ -412,30 +429,3 @@ are the same for all three:
|
|
|
412
429
|
## Cleaning up
|
|
413
430
|
|
|
414
431
|
`delete_creative` and `delete_media` are permanent. Never call them unless the person asked or confirmed that exact item; the server refuses a Creative still on a Post or with other variants, and Media still in use.
|
|
415
|
-
|
|
416
|
-
## Failures that are silent (carried over from this skill's predecessor)
|
|
417
|
-
|
|
418
|
-
Automated checks are real but incomplete. These are the ones worth remembering to look for by eye:
|
|
419
|
-
|
|
420
|
-
- **A photo with text, a logo or a claim burned into its pixels** passes every copy-rule and
|
|
421
|
-
render check, because none of them look at pixels for meaning — only step 5's "open and look at
|
|
422
|
-
it" catches this.
|
|
423
|
-
- **A generated image depicting a product, result or claim that isn't real** (a made-up app screen,
|
|
424
|
-
a before/after, an invented package) passes every check — only looking at it catches it.
|
|
425
|
-
- **A structurally valid render is not necessarily a good one.** The render's own assertions check
|
|
426
|
-
spill, safe zones, fonts and contrast — never whether the headline reads at a glance or the copy
|
|
427
|
-
sits on the subject's face. Step 7 exists because of this gap.
|
|
428
|
-
- **A Brand Kit with no copy rules, or no accent-phrase requirement, never fails a save on that
|
|
429
|
-
account.** That is permissive by design, not a bug — but it means the model, not the platform, is
|
|
430
|
-
the only check on brand voice for that workspace until someone configures rules.
|
|
431
|
-
- **A draft is not a saved Creative.** `preview_creative_draft` writes nothing to Conductor, so a
|
|
432
|
-
Creative or media id does not exist until `commit_creative_draft` runs — don't hand a draft's
|
|
433
|
-
folder to `attach_creative_to_post` or a Post.
|
|
434
|
-
- **A stale write is refused, not silently overwritten.** `update_creative` requires the Creative's
|
|
435
|
-
current `version`; a mismatch comes back as a 409 naming the conflict rather than clobbering a
|
|
436
|
-
concurrent edit — reread with `get_creative` and reapply, don't retry blind.
|
|
437
|
-
|
|
438
|
-
## Related skills
|
|
439
|
-
|
|
440
|
-
- `conductor-publisher`: schedules and tracks the Post this Creative's artwork attaches to, through
|
|
441
|
-
review and out to each destination. This skill hands off to it, never the reverse.
|
package/assets/claude/skills/{conductor-creative → conductor-marketing-create}/references/hooks.md
RENAMED
|
@@ -1,4 +1,4 @@
|
|
|
1
|
-
<!-- Used by conductor-
|
|
1
|
+
<!-- Used by the Direct mode of conductor-marketing-create. -->
|
|
2
2
|
# Hooks: writing the headline
|
|
3
3
|
|
|
4
4
|
A hook is the line that makes a thumb stop. A tagline says what the brand is; a hook names a moment,
|
|
@@ -8,12 +8,14 @@ an object or a number the viewer recognises. Write several, score them, keep the
|
|
|
8
8
|
|
|
9
9
|
Mix the types. Do not write ten versions of one.
|
|
10
10
|
|
|
11
|
-
- **Situation / POV**: the moment the viewer is in. "
|
|
12
|
-
- **Mechanism**: how it works, in one line. "
|
|
11
|
+
- **Situation / POV**: the moment the viewer is in. "which tab had the quote?"
|
|
12
|
+
- **Mechanism**: how it works, in one line. "paste a link. skip the rereading."
|
|
13
13
|
- **Enemy**: the thing it replaces or kills. "The spreadsheet nobody updates."
|
|
14
|
+
Mechanism and Enemy hooks need an attitude to pass `voice.md`'s self-check: "the spreadsheet nobody
|
|
15
|
+
updates" has one; a bare feature label ("Paste a link, get a summary") does not.
|
|
14
16
|
- **Exact result**: a specific number or time. "Invoices sent in 4 minutes, not 4 days."
|
|
15
17
|
- **Concept**: a visual metaphor the image carries, with the line finishing the thought. A photo of one
|
|
16
|
-
chair at a long empty table, "
|
|
18
|
+
chair at a long empty table, "this chair has been waiting since 9:15".
|
|
17
19
|
|
|
18
20
|
An approved line from the Knowledge page still wins when one fits. Quote it verbatim; use the
|
|
19
21
|
candidates to find the angle around it or when none fits.
|
|
@@ -26,6 +28,8 @@ promise that could open any account's post, 5 a specific, surprising line a stra
|
|
|
26
28
|
- **Hook strength**: would it stop a thumb?
|
|
27
29
|
- **Specificity**: does it name a moment, an object or a number?
|
|
28
30
|
- **Competitor test**: could a competitor run this unchanged? If yes, it fails; rewrite or drop it.
|
|
31
|
+
- **Reads like a person**: would someone say it out loud to a friend? Run `voice.md`'s self-check;
|
|
32
|
+
a template frame ("X but Y"), a brand word or a line that describes the photo fails.
|
|
29
33
|
|
|
30
34
|
Exact numbers read as more credible than round ones: "11 minutes" over "10 minutes". One selling point
|
|
31
35
|
per line, written the way a person would say it.
|
|
@@ -39,9 +43,9 @@ hooks for variants (`variantOf`) or the caption. Run the chosen line through the
|
|
|
39
43
|
|
|
40
44
|
| Tagline | Hook |
|
|
41
45
|
|---|---|
|
|
42
|
-
| Smarter scheduling for busy teams |
|
|
43
|
-
| Fresh food, delivered |
|
|
44
|
-
| Save money on energy |
|
|
45
|
-
| Learn faster with AI |
|
|
46
|
+
| Smarter scheduling for busy teams | this meeting could've been a link |
|
|
47
|
+
| Fresh food, delivered | the oven's still cold. dinner's here. |
|
|
48
|
+
| Save money on energy | our own bill dropped $38 last month |
|
|
49
|
+
| Learn faster with AI | ordered lunch in Spanish on day 30 |
|
|
46
50
|
|
|
47
|
-
Check the
|
|
51
|
+
Check the number and speed rows against the brand's claims rules before using a line like them: a number or a result has to be true.
|
|
@@ -11,8 +11,8 @@ Heading: `## Idea <n>: <title>` where the title is 8 words or fewer.
|
|
|
11
11
|
|
|
12
12
|
- **Concept:** <one line, max 25 words: what the viewer sees and why it works>
|
|
13
13
|
- **Hook, spoken:** <max 12 words, or "none (text only)">
|
|
14
|
-
- **Hook, on screen:** <max 12 words
|
|
15
|
-
- **Hook type:** <question | bold-claim | pattern-interrupt | relatable-pain | how-to | number | curiosity-gap | story | social-proof>
|
|
14
|
+
- **Hook, on screen:** <max 12 words; on the artwork aim for about 8 or fewer, `voice.md`>
|
|
15
|
+
- **Hook type:** <question | bold-claim | pattern-interrupt | relatable-pain | how-to | number | curiosity-gap | story | social-proof; tag it from `voice.md`'s "Shapes, and the hookType each one is tagged" table>
|
|
16
16
|
- **Format:** <slideshow | still | carousel | story sequence | motion text over photo | motion text over clip | talking head | demo | skit | other>
|
|
17
17
|
- **Platform:** <primary platform, then any others>
|
|
18
18
|
- **Tag:** RENDER | FILM
|
|
@@ -1,4 +1,3 @@
|
|
|
1
|
-
<!-- Keep identical to the copy in the other creative skill (conductor-creative / conductor-content-studio references/layout-guide.md). Skills install independently and cannot reference each other. -->
|
|
2
1
|
# Layout guide
|
|
3
2
|
|
|
4
3
|
Pick the layout on purpose. A Creative has one layout across all its placements (9:16, 4:5, 1:1), so
|
|
@@ -1,4 +1,3 @@
|
|
|
1
|
-
<!-- Keep identical to the copy in the other creative skill (conductor-creative / conductor-content-studio references/realism-check.md). Skills install independently and cannot reference each other. -->
|
|
2
1
|
# Make it real, then check it
|
|
3
2
|
|
|
4
3
|
This is about image quality, not concealment. The `aiGenerated: true` provenance and the platform's
|
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
# Judging rubric
|
|
2
2
|
|
|
3
|
-
Used by the `conductor-
|
|
3
|
+
Used by the `conductor-marketing-creative-director` agent in `text-round` and `visual-round`, and by
|
|
4
4
|
the orchestrator when it needs to break a tie. Everything here exists to counter a known judge
|
|
5
5
|
failure mode, so follow it as written.
|
|
6
6
|
|
|
@@ -23,11 +23,12 @@ Score each 1 to 5. Weighted total = sum(score x weight) / 5, so the maximum is 1
|
|
|
23
23
|
|
|
24
24
|
| Criterion | Weight | 1 | 3 | 5 |
|
|
25
25
|
|---|---|---|---|---|
|
|
26
|
-
| Hook strength |
|
|
26
|
+
| Hook strength | 20 | Opens on a greeting, a logo or context; nothing makes a stranger stay | A clear promise or question, but generic; could open any account's post | A specific, surprising or tension-filled claim a stranger understands in 1s, spoken and on screen |
|
|
27
27
|
| Audience relevance | 20 | Aimed at "everyone"; ignores the brief's tension | Names the audience's topic but not their actual friction | Opens on a friction the brief names and the audience would recognise as theirs |
|
|
28
|
-
|
|
|
28
|
+
| Reads like a person | 15 | A template frame ("X but Y"), brand words, a described picture or an explained joke; nobody would say it out loud | Plain and correct, but no speaker or attitude; reads as a caption, not a post | Someone is talking, with one real detail and a reason to care; passes every question in `voice.md`'s self-check |
|
|
29
|
+
| Brand fit | 10 | Breaks a copy rule or approved-line, or reads off-voice | Within the rules but could be any brand | Uses an approved line or the knowledge page's voice naturally; no rule risk |
|
|
29
30
|
| Feasibility with assets on hand | 15 | Needs assets that do not exist and cannot be made here (and is tagged `RENDER`) | Needs one new photo or clip the user can supply easily | Fully buildable from `media.json` or a render with no photo; or an honest `FILM` with a simple shoot |
|
|
30
|
-
| Novelty |
|
|
31
|
+
| Novelty | 10 | The mechanic is the most common one for this topic (listicle, "5 tips", before/after) | A familiar mechanic with a fresh angle | A mechanic not seen elsewhere in the pool or in the performance data, still easy to grasp. Novelty is in the idea, not in an odd sentence: a strange word pairing scores nothing here |
|
|
31
32
|
| Shareability | 10 | Nothing to send to a friend; purely informational | Useful or pleasant, but no reason to send | A person would send it to a named other person ("this is you") or save it for later |
|
|
32
33
|
|
|
33
34
|
## Scoring protocol
|
|
@@ -37,13 +38,19 @@ Score each 1 to 5. Weighted total = sum(score x weight) / 5, so the maximum is 1
|
|
|
37
38
|
2. For each idea, per criterion, write one or two sentences of reasoning FIRST, citing the idea's
|
|
38
39
|
own words. Then write the score. Never write a score before its reasoning.
|
|
39
40
|
3. Penalise nothing for length (ideas are length-matched) and reward nothing for confident tone.
|
|
40
|
-
4. Compute the weighted total.
|
|
41
|
+
4. Compute the weighted total.
|
|
41
42
|
5. Any idea that breaks a copy rule or contradicts an approved line is capped at 2 on brand fit
|
|
42
43
|
and cannot be a finalist unless the fix is a one-word edit you state.
|
|
44
|
+
6. Any idea scoring 1 or 2 on reads-like-a-person cannot be a finalist, full stop. Do not rewrite
|
|
45
|
+
it and score the rewrite.
|
|
46
|
+
7. Shortlist the top 3 eligible ideas (steps 5 and 6 applied first). If ties, apply the tie-break
|
|
47
|
+
rule below. If fewer than 3 are eligible, say so in `scores.md` and continue with what remains:
|
|
48
|
+
the pairwise round covers the remaining pairs, and a single eligible idea wins outright (the
|
|
49
|
+
visual round only checks it against the veto questions).
|
|
43
50
|
|
|
44
51
|
## Pairwise protocol (visual round)
|
|
45
52
|
|
|
46
|
-
For each pair of finalists (A,B), (A,C), (B,C):
|
|
53
|
+
For each pair of finalists (for three: (A,B), (A,C), (B,C); for two, the one pair; for one, skip this):
|
|
47
54
|
|
|
48
55
|
1. Judge the pair twice, once with each order, as two independent comparisons. Write reasoning
|
|
49
56
|
first, then the verdict "first" or "second", then say which idea won.
|
|
@@ -65,6 +72,8 @@ Look at the actual image file, not the text description. Check:
|
|
|
65
72
|
payoff not the opener, native look, not a banner ad), each 1 to 5. A finalist with a criterion below 3
|
|
66
73
|
that the layout could have fixed (`stacked` or `split` on 9:16, a dead panel) counts against it.
|
|
67
74
|
- Realism: a generated photo that looks AI-made (waxy, duplicated objects, warped hands) counts against the finalist.
|
|
75
|
+
- Voice: read every card and the caption against `voice.md`'s self-check. A card that describes
|
|
76
|
+
the picture, a template frame or a brand word counts against the finalist.
|
|
68
77
|
- Whether the picture and the words do different jobs (photo shows, words claim) rather than
|
|
69
78
|
repeating each other.
|
|
70
79
|
|
|
@@ -12,7 +12,7 @@ Idea: <id> Source decision: decision.md
|
|
|
12
12
|
|
|
13
13
|
## Hook (0-3s)
|
|
14
14
|
- Spoken: <line, max 12 words, or "none">
|
|
15
|
-
- On screen: <text, max 12 words>
|
|
15
|
+
- On screen: <text, max 12 words; on the artwork about 8 or fewer>
|
|
16
16
|
- Visual: <what is shown>
|
|
17
17
|
|
|
18
18
|
## Beats
|
|
@@ -42,10 +42,12 @@ opening words (and in the on-screen text and speech); the first line carries the
|
|
|
42
42
|
3. <optional>
|
|
43
43
|
|
|
44
44
|
Hook types: question, bold-claim, pattern-interrupt, relatable-pain, how-to, number,
|
|
45
|
-
curiosity-gap, story, social-proof.
|
|
45
|
+
curiosity-gap, story, social-proof (tag from `voice.md`'s shapes table).
|
|
46
46
|
|
|
47
47
|
## Copy-rule check
|
|
48
48
|
<each must-say / must-avoid item from brief.md, and ok or the fix>
|
|
49
|
+
## Voice check
|
|
50
|
+
<each line changed against voice.md, the hook included: before, after, which question or tell it failed; or "all lines pass". Functional lines (CTA, save or follow ask, link in bio, pinned comment, plain app mention, disclosure, approved lines) are checked only against "Tells to cut". Wording the person asked for stays; note any tell it trips here.>
|
|
49
51
|
```
|
|
50
52
|
|
|
51
53
|
Rules: a beat every 3 to 5 seconds (a `MOTION` idea instead has one headline, one optional body and an end card, with no timed on-screen beats); no beat longer than 5 seconds without a visual or text change;
|
|
@@ -1,4 +1,3 @@
|
|
|
1
|
-
<!-- Keep identical to the copy in the other creative skill (conductor-creative / conductor-content-studio references/slideshow-formats.md). Skills install independently and cannot reference each other. -->
|
|
2
1
|
# Slideshow formats
|
|
3
2
|
|
|
4
3
|
For organic posts: a native-looking photo slideshow (TikTok photo post, Instagram carousel) that
|
|
@@ -70,22 +69,26 @@ A five-slide "stop doing X" post. Paths are relative to the spec file or absolut
|
|
|
70
69
|
## Hook archetypes
|
|
71
70
|
|
|
72
71
|
Pick one per post and name it in the idea. Change one thing per post so results are comparable.
|
|
72
|
+
The slide-1 examples are shapes only: each must pass `voice.md`'s self-check with your own true
|
|
73
|
+
detail, and none may state a fact the brief does not support.
|
|
73
74
|
|
|
74
75
|
| Archetype | Slide 1 | Slides |
|
|
75
76
|
|---|---|---|
|
|
76
|
-
| Saved-but-never-used guilt | "
|
|
77
|
-
| Doubt or conflict |
|
|
78
|
-
| Stop doing X | "
|
|
79
|
-
| Numbered list | "5
|
|
80
|
-
| Tier list | "
|
|
81
|
-
| Notes-app plan | "
|
|
82
|
-
| POV | "
|
|
83
|
-
| Before / after | "
|
|
77
|
+
| Saved-but-never-used guilt | "the recipes folder is a graveyard" | `grid` (one tile `restore`d at the end), `photo`, `screen`, ask |
|
|
78
|
+
| Doubt or conflict | A real message someone sent you (needs a true story) | `messages`, `photo`, `notes`, `screen`, ask |
|
|
79
|
+
| Stop doing X | "stop checking your balance at midnight" | `photo` with the reason per slide, `screen`, ask |
|
|
80
|
+
| Numbered list | "5 stretches for desk-bound shoulders" | `photo` per item, `screen` last, ask |
|
|
81
|
+
| Tier list | "ranking airport snacks. no appeals" | `tier`, `photo` close-ups, `screen`, ask |
|
|
82
|
+
| Notes-app plan | "my monday plan, aka the optimistic draft" | `notes`, `week`, `screen`, ask |
|
|
83
|
+
| POV | "pov: it's sunday and the week is meal-prepped" | `photo` (hands or a real scene), `screen`, `photo`, ask |
|
|
84
|
+
| Before / after | "week 1 vs week 4, same grocery run" (needs real before and after) | `stack` with `label`s, `photo`, `screen`, ask |
|
|
84
85
|
|
|
85
86
|
## Text rules
|
|
86
87
|
|
|
87
88
|
- Native box text (`box`) or white text with a black stroke (`stroke`); centred, bold, sans.
|
|
88
89
|
- At most 8 words a line and 3 lines a box.
|
|
90
|
+
- Every card reads like a person posted it (`voice.md`): a speaker, one real detail, no template
|
|
91
|
+
frames, never a description of the photo.
|
|
89
92
|
- Keep text between about y 16 and y 75 (percent of the 9:16 frame). That survives the 4:5 Instagram
|
|
90
93
|
cut and the platform UI (top bar, bottom caption, right button column).
|
|
91
94
|
- The first text of every slide needs `y`.
|
|
@@ -1,15 +1,9 @@
|
|
|
1
|
-
|
|
2
|
-
|
|
3
|
-
|
|
4
|
-
|
|
5
|
-
|
|
6
|
-
|
|
7
|
-
|
|
8
|
-
# Conductor Content Studio
|
|
9
|
-
|
|
10
|
-
Takes a request ("a post about X for Y") to either a committed Creative or a shoot-ready script. The
|
|
11
|
-
team chooses the idea itself; the person is asked once, at the final draft preview. You are the
|
|
12
|
-
orchestrator and run in the main session. Subagents do the thinking and writing; you do every
|
|
1
|
+
# Team mode: from a goal to a committed Creative
|
|
2
|
+
|
|
3
|
+
Read when the person gives a goal ("a post about X for Y") rather than a headline, photo or clip. The
|
|
4
|
+
shared rules in `SKILL.md` apply. Takes the request to either a committed Creative or a shoot-ready
|
|
5
|
+
script. The team chooses the idea itself; the person is asked once, at the final draft preview. You
|
|
6
|
+
are the orchestrator and run in the main session. Subagents do the thinking and writing; you do every
|
|
13
7
|
Conductor call and every file merge.
|
|
14
8
|
|
|
15
9
|
## Rules that shape the design
|
|
@@ -46,7 +40,8 @@ Each exists because of a known failure mode (sources are in `references/rubric.m
|
|
|
46
40
|
rules into `context/banned.md` on every later run.
|
|
47
41
|
|
|
48
42
|
Cost: a run is about 8 subagent calls (strategist 1, ideators 3, director 2, scriptwriter 1, art
|
|
49
|
-
director 1)
|
|
43
|
+
director 1). The strategist, the director and the scriptwriter use the top model tier; the
|
|
44
|
+
ideators stay on the lower tier so the judge never scores its own tier's ideas. `--quick` runs two
|
|
50
45
|
ideators and skips the finalist renders (about 7 calls).
|
|
51
46
|
|
|
52
47
|
## Run folder and state
|
|
@@ -107,7 +102,7 @@ template exactly; <any stage-specific rule>
|
|
|
107
102
|
Return: at most 150 words plus the output path.
|
|
108
103
|
```
|
|
109
104
|
|
|
110
|
-
Resolve the reference files once: Glob for `**/conductor-
|
|
105
|
+
Resolve the reference files once: Glob for `**/conductor-marketing-create/references/rubric.md` under
|
|
111
106
|
`.claude/skills/` in this project, then under `~/.claude/skills/`. Use that folder for every
|
|
112
107
|
`references/` path below. A subagent that returns more than 150 words or leaves its file unwritten
|
|
113
108
|
has failed the stage: retry once with a one-line reminder, then tell the person.
|
|
@@ -161,20 +156,20 @@ Write `request.md` (the request verbatim, `--quick` if given, the platform if st
|
|
|
161
156
|
|
|
162
157
|
### S1 — Strategist -> `brief.md`
|
|
163
158
|
|
|
164
|
-
`Agent(subagent_type: "conductor-
|
|
159
|
+
`Agent(subagent_type: "conductor-marketing-strategist")` with the seven context files (including `context/banned.md`, whose every entry goes into the brief's must-avoid), `request.md`
|
|
165
160
|
and `references/capabilities.md` (so the brief states rendering constraints correctly: motion text
|
|
166
161
|
over a photo is renderable and needs no filming).
|
|
167
162
|
|
|
168
163
|
### S2 — Ideators, in parallel -> `ideas/angle-<n>.md`
|
|
169
164
|
|
|
170
165
|
Three `Agent` calls in ONE message so they run in parallel, `subagent_type:
|
|
171
|
-
"conductor-
|
|
166
|
+
"conductor-marketing-ideator"`, with these angles and no knowledge of each other:
|
|
172
167
|
|
|
173
168
|
1. story/emotion 2. utility/education 3. trend/humor (the only one allowed web search)
|
|
174
169
|
|
|
175
|
-
Each gets `brief.md`, `context/media.json`, `references/idea-template.md
|
|
176
|
-
`references/capabilities.md` (the `RENDER`/`FILM` test)
|
|
177
|
-
With `--quick`: only angles 1 and 2, 3 ideas each.
|
|
170
|
+
Each gets `brief.md`, `context/media.json`, `references/idea-template.md`,
|
|
171
|
+
`references/capabilities.md` (the `RENDER`/`FILM` test) and `references/voice.md`, and writes 4 ideas.
|
|
172
|
+
With `--quick`: only angles 1 and 2, 3 ideas each. Any ideator rerun passes the same inputs, `references/voice.md` included.
|
|
178
173
|
|
|
179
174
|
### Merge (you) -> `ideas/pool.md`
|
|
180
175
|
|
|
@@ -189,12 +184,14 @@ Build one pool from all the angle files:
|
|
|
189
184
|
|
|
190
185
|
### S3a — Creative director, text round -> `judging/scores.md`
|
|
191
186
|
|
|
192
|
-
`Agent(subagent_type: "conductor-
|
|
193
|
-
`brief.md`, `ideas/pool.md`, `references/rubric.md`. It scores every idea independently, reasoning
|
|
194
|
-
before each score, and names the top 3. Store the finalists in `run.json`.
|
|
187
|
+
`Agent(subagent_type: "conductor-marketing-creative-director")`, mode `text-round`. Inputs:
|
|
188
|
+
`brief.md`, `ideas/pool.md`, `references/rubric.md`, `references/voice.md`. It scores every idea independently, reasoning
|
|
189
|
+
before each score, and names the top 3 eligible ideas (fewer when the rubric's gates leave fewer; it says so). Store the finalists in `run.json`.
|
|
195
190
|
|
|
196
191
|
### S3b — Finalist drafts (you; skipped in `--quick`)
|
|
197
192
|
|
|
193
|
+
Draft each finalist the judge named (up to three).
|
|
194
|
+
|
|
198
195
|
A slideshow finalist has no cheap draft: judge it from its idea text (archetype and slide-by-slide
|
|
199
196
|
text) unless its slides can be composed from photos already on hand. In that case write its spec to
|
|
200
197
|
`<run>/drafts/finalist-<id>/slideshow-spec.json`, call `compose_slideshow` with `draftDir` set to
|
|
@@ -219,7 +216,7 @@ path as `photoPath` (generate only for finalists, not the whole pool).
|
|
|
219
216
|
|
|
220
217
|
Same agent, mode `visual-round`. Inputs: `brief.md`, `judging/scores.md`, `ideas/pool.md`, the
|
|
221
218
|
finalist image paths (`drafts/finalist-<id>/sheet.jpg`; the director views them with Read), the
|
|
222
|
-
rubric, `references/design-critique.md` and `references/compliance-checklist.md`. It compares finalists pairwise in both orders, runs
|
|
219
|
+
rubric, `references/voice.md`, `references/design-critique.md` and `references/compliance-checklist.md`. It compares finalists pairwise in both orders (a single finalist wins outright and is only checked), runs
|
|
223
220
|
the devil's-advocate pass against the leader (including the veto questions: an unsubstantiated
|
|
224
221
|
claim, trademark or parody risk, a representation or sensitivity problem, disclosure needed), and
|
|
225
222
|
writes `decision.md`: the winner, why it won, the runners-up, one line per cut idea and the veto
|
|
@@ -234,8 +231,8 @@ which idea to take; show the decision at the gate.
|
|
|
234
231
|
### S4 — Scriptwriter -> `script.md`
|
|
235
232
|
|
|
236
233
|
First copy the winning idea's block from `ideas/pool.md` to `ideas/winner.md`. Then
|
|
237
|
-
`Agent(subagent_type: "conductor-
|
|
238
|
-
`references/publish-pack-template.md
|
|
234
|
+
`Agent(subagent_type: "conductor-marketing-scriptwriter")` with `brief.md`, `decision.md`, `ideas/winner.md`, `references/script-template.md`, `references/platform-specs.md`,
|
|
235
|
+
`references/publish-pack-template.md` and `references/voice.md` (required on every scriptwriter run, including the reruns below). Besides the script it writes the "Numbers and sources" list (every number on a card or in a caption, with its source record) and the Publish copy section: a
|
|
239
236
|
caption per platform whose call to action names the app so it can be searched, hashtags, a pinned
|
|
240
237
|
comment and 2 to 3 alternate hooks with a hookType each.
|
|
241
238
|
For a `MOTION` winner tell it that the render shows one headline and one optional body, animated
|
|
@@ -254,7 +251,7 @@ through `generate_creative_image` and the realism check first, as below. Every p
|
|
|
254
251
|
through the resolution gate below. The rest of this stage is
|
|
255
252
|
the STILL/MOTION path.
|
|
256
253
|
|
|
257
|
-
`Agent(subagent_type: "conductor-
|
|
254
|
+
`Agent(subagent_type: "conductor-marketing-art-director")` with `script.md`, `context/media.json`,
|
|
258
255
|
`context/brand-kit.json`, `context/banned.md`, `references/shotlist-template.md`,
|
|
259
256
|
`references/platform-specs.md`, `references/realism-check.md` and
|
|
260
257
|
`references/publish-pack-template.md`. Besides the spec it writes the Publish assets section of
|
|
@@ -304,7 +301,7 @@ does not apply is answered "not applicable".
|
|
|
304
301
|
sets `brandContentToggle`; gifted or affiliate sets `brandContentToggle` and adds the
|
|
305
302
|
disclosure to the caption's first line.
|
|
306
303
|
4. A claims or trademark check that fails and cannot be fixed by you in one word goes back to the
|
|
307
|
-
scriptwriter once, with the failing line
|
|
304
|
+
scriptwriter once, with the failing line and `references/voice.md`; re-run S4's merge and S5 only if the copy changed. If
|
|
308
305
|
it still fails, show it at the gate as a known issue and never hide it.
|
|
309
306
|
5. Fact lint: merge the scriptwriter's "Numbers and sources" into the pack and check each number
|
|
310
307
|
against its source record (checklist item 9). Fix a mismatch before the gate.
|
|
@@ -338,7 +335,7 @@ any placement is fixed now, not shown.
|
|
|
338
335
|
`audio.source: "track"` previews as its slideshow MP4 too (`slideshowVideoPath`), so give the
|
|
339
336
|
person that path to watch along with the sheet, and put the track's licence in the pack's music
|
|
340
337
|
line (`references/compliance-checklist.md` section 3).
|
|
341
|
-
2. Eye-check the image.
|
|
338
|
+
2. Eye-check the image. Read `references/direct.md` (same skill) and
|
|
342
339
|
apply its review guidance (step 7 and the silent-failures list), scoring the draft with
|
|
343
340
|
`references/design-critique.md` (the slideshow criteria for a slideshow, the STILL criteria
|
|
344
341
|
otherwise; any criterion below 3 is fixed before anyone sees it). For a STILL: does the headline read at a
|
|
@@ -369,10 +366,10 @@ any placement is fixed now, not shown.
|
|
|
369
366
|
its state into `publish-pack.json` as `creativeState` (and both into `run.json`).
|
|
370
367
|
- **Use a runner-up**: ask a second `AskUserQuestion` listing the runners-up by title (with
|
|
371
368
|
their tag; label one that could not be rendered as such and say why, rather than offering it as
|
|
372
|
-
if it would work). Then set the chosen idea as `winner` and re-run S4 onward for it
|
|
373
|
-
or the FILM path if it is tagged `FILM`).
|
|
369
|
+
if it would work). Then set the chosen idea as `winner` and re-run S4 onward for it, passing
|
|
370
|
+
`references/voice.md` again (S5, S5b and S6, or the FILM path if it is tagged `FILM`).
|
|
374
371
|
- **Request changes**: take the free text. If it is about the copy, send it with the current
|
|
375
|
-
script to the scriptwriter (S4), then the art director (S5); if only about the look, send it to
|
|
372
|
+
script and `references/voice.md` to the scriptwriter (S4), then the art director (S5); if only about the look, send it to
|
|
376
373
|
the art director. Redo S5b, then re-preview. Do not commit the old draft.
|
|
377
374
|
- **Stop**: upload nothing. Tell the person where the drafts are
|
|
378
375
|
(`drafts.final` and the finalist folders) and that the run can be resumed.
|
|
@@ -380,7 +377,7 @@ any placement is fixed now, not shown.
|
|
|
380
377
|
### FILM winner
|
|
381
378
|
|
|
382
379
|
1. After S4, have the art director write the shoot version: `Agent(subagent_type:
|
|
383
|
-
"conductor-
|
|
380
|
+
"conductor-marketing-art-director")` with `script.md`, `context/media.json`, `context/brand-kit.json`
|
|
384
381
|
and `references/shotlist-template.md`, output `shoot/shoot-script.md`: script, shot list, capture
|
|
385
382
|
settings (9:16, 30 fps, sound on), B-roll, safe zones, and the on-screen text and captions ready to
|
|
386
383
|
paste. It also writes the Publish assets section there. It writes no `creative-spec.json` here.
|
|
@@ -403,11 +400,10 @@ Keep it short:
|
|
|
403
400
|
alternate hooks); say the publisher will pre-fill the Post from `publish-pack.json`, and that the
|
|
404
401
|
pinned comment, the link in bio and the Stories repost with a link sticker are manual steps;
|
|
405
402
|
- next steps:
|
|
406
|
-
- make the Post with
|
|
403
|
+
- make the Post with the `conductor-marketing-publish` skill (the Creative is
|
|
407
404
|
not on a Post yet, because a run never creates one);
|
|
408
405
|
- to test a runner-up's hook, cut it as a lettered variant of the Creative with `variantOf`;
|
|
409
|
-
- for `FILM`: film it, upload the clip, then run `/conductor:
|
|
410
|
-
clip I have".
|
|
406
|
+
- for `FILM`: film it, upload the clip, then run `/conductor:marketing create` and pick a clip I have.
|
|
411
407
|
|
|
412
408
|
## Failure handling
|
|
413
409
|
|
|
@@ -421,6 +417,6 @@ Keep it short:
|
|
|
421
417
|
|
|
422
418
|
## Related
|
|
423
419
|
|
|
424
|
-
- `
|
|
420
|
+
- `references/direct.md`: the single-Creative flow whose checklist S6 uses.
|
|
425
421
|
- `references/slideshow-formats.md`: the slideshow spec, archetypes and text rules.
|
|
426
|
-
- `conductor-
|
|
422
|
+
- `conductor-marketing-publish`: takes the committed Creative onto a Post, pre-filled from `publish-pack.json`.
|