@cliangdev/conductor 0.41.1 → 0.42.0

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.
Files changed (43) hide show
  1. package/assets/claude/agents/conductor-coder.md +278 -7
  2. package/assets/claude/agents/{conductor-content-art-director.md → conductor-marketing-art-director.md} +2 -2
  3. package/assets/claude/agents/{conductor-content-creative-director.md → conductor-marketing-creative-director.md} +2 -2
  4. package/assets/claude/agents/{conductor-content-ideator.md → conductor-marketing-ideator.md} +3 -3
  5. package/assets/claude/agents/{conductor-content-scriptwriter.md → conductor-marketing-scriptwriter.md} +2 -2
  6. package/assets/claude/agents/{conductor-content-strategist.md → conductor-marketing-strategist.md} +2 -2
  7. package/assets/claude/commands/conductor/implement.md +1 -1
  8. package/assets/claude/commands/conductor/marketing.md +26 -0
  9. package/assets/claude/skills/conductor-marketing-create/SKILL.md +85 -0
  10. package/assets/claude/skills/{conductor-creative → conductor-marketing-create}/references/brand-kit-setup.md +2 -3
  11. package/assets/claude/skills/{conductor-content-studio → conductor-marketing-create}/references/design-critique.md +0 -1
  12. package/assets/claude/skills/{conductor-creative/SKILL.md → conductor-marketing-create/references/direct.md} +48 -59
  13. package/assets/claude/skills/{conductor-creative → conductor-marketing-create}/references/hooks.md +1 -1
  14. package/assets/claude/skills/{conductor-creative → conductor-marketing-create}/references/layout-guide.md +0 -1
  15. package/assets/claude/skills/{conductor-creative → conductor-marketing-create}/references/realism-check.md +0 -1
  16. package/assets/claude/skills/{conductor-content-studio → conductor-marketing-create}/references/rubric.md +1 -1
  17. package/assets/claude/skills/{conductor-creative → conductor-marketing-create}/references/slideshow-formats.md +0 -1
  18. package/assets/claude/skills/{conductor-content-studio/SKILL.md → conductor-marketing-create/references/team-run.md} +18 -25
  19. package/assets/claude/skills/{conductor-publisher → conductor-marketing-publish}/SKILL.md +5 -5
  20. package/assets/claude/skills/{conductor-marketing-analyst → conductor-marketing-report}/SKILL.md +11 -4
  21. package/assets/cli-manifest.json +17 -37
  22. package/dist/lib/plugin-assets.d.ts.map +1 -1
  23. package/dist/lib/plugin-assets.js +97 -42
  24. package/dist/lib/plugin-assets.js.map +1 -1
  25. package/package.json +1 -1
  26. package/assets/claude/commands/conductor/content.md +0 -66
  27. package/assets/claude/commands/conductor/creative.md +0 -108
  28. package/assets/claude/commands/conductor/marketing-report.md +0 -38
  29. package/assets/claude/skills/conductor-coder/SKILL.md +0 -282
  30. package/assets/claude/skills/conductor-content-studio/references/layout-guide.md +0 -73
  31. package/assets/claude/skills/conductor-content-studio/references/realism-check.md +0 -95
  32. package/assets/claude/skills/conductor-content-studio/references/slideshow-formats.md +0 -135
  33. package/assets/claude/skills/conductor-creative/references/design-critique.md +0 -52
  34. package/assets/claude/skills/conductor-ux-ui-design/SKILL.md +0 -346
  35. package/assets/claude/skills/conductor-ux-ui-design/references/design-tokens.md +0 -359
  36. /package/assets/claude/skills/{conductor-content-studio → conductor-marketing-create}/references/capabilities.md +0 -0
  37. /package/assets/claude/skills/{conductor-content-studio → conductor-marketing-create}/references/compliance-checklist.md +0 -0
  38. /package/assets/claude/skills/{conductor-content-studio → conductor-marketing-create}/references/idea-template.md +0 -0
  39. /package/assets/claude/skills/{conductor-content-studio → conductor-marketing-create}/references/platform-specs.md +0 -0
  40. /package/assets/claude/skills/{conductor-content-studio → conductor-marketing-create}/references/publish-pack-template.md +0 -0
  41. /package/assets/claude/skills/{conductor-content-studio → conductor-marketing-create}/references/script-template.md +0 -0
  42. /package/assets/claude/skills/{conductor-content-studio → conductor-marketing-create}/references/shotlist-template.md +0 -0
  43. /package/assets/claude/skills/{conductor-marketing-analyst → conductor-marketing-report}/references/html-layout.md +0 -0
@@ -1,66 +0,0 @@
1
- ---
2
- name: conductor:content
3
- description: Run a small team of agents that brainstorm competing short-form social post ideas, pick one with a bias-resistant judging method, write the script and visual direction, and render a local draft. You approve at the preview; only then is the Creative uploaded to Conductor. Never creates a Post.
4
- argument-hint: "[request] [--quick]"
5
- allowed-tools: mcp__conductor__*, Agent, AskUserQuestion, Skill, Read, Write, Glob, Grep, WebFetch
6
- ---
7
-
8
- # /conductor:content
9
-
10
- One command from "I need a post for X" to a committed Creative (or, for an idea that needs filming,
11
- a shoot-ready script). The team chooses the idea itself; the only point where it stops for you is the
12
- final draft preview. This command gathers the request and hands off to the `conductor-content-studio`
13
- skill (`Skill(skill: "conductor-content-studio")`), which runs the stages. Don't re-derive what that
14
- skill covers. For an organic post the team's renderable ideas are native slideshows (built with
15
- `compose_slideshow`, posted as a CAROUSEL Creative); stills and motion pieces stay for ideas that
16
- are one image, an ad, or a motion piece.
17
-
18
- ## Trigger
19
-
20
- `/conductor:content "<request>"`, optionally with `--quick`, or a request to have a team come up
21
- with and make a short-form social post.
22
-
23
- - The request is free text: what the post is for, the platform, any line it must say.
24
- - `--quick` runs a lighter version: two ideators, no finalist drafts rendered before the judge
25
- decides. Pass it through to the skill.
26
-
27
- ## Step 1 — Check the brand
28
-
29
- Read `references/brand-kit-setup.md` in the `conductor-creative` skill
30
- (`.claude/skills/conductor-creative/references/brand-kit-setup.md`, in this project or under `~/.claude`)
31
- and follow it: check the Brand Kit, and offer to set it up if it has never been configured. Then
32
- continue with Step 2. The team writes copy against the kit's rules and approved lines, so an
33
- unconfigured kit makes for generic ideas; say so if the person skips setup.
34
-
35
- ## Step 2 — The request
36
-
37
- If a request came with the command, keep it as given and go to Step 3.
38
-
39
- If not, ask one short question with `AskUserQuestion` (one batch, not an interview):
40
-
41
- ```json
42
- {
43
- "questions": [{
44
- "question": "What should the post be for? Include the platform and any line it must say.",
45
- "header": "Request",
46
- "options": [
47
- {"label": "Describe it", "description": "Free text: the goal or topic, the platform (TikTok, Reels, Shorts, Facebook), any must-say line"},
48
- {"label": "Surprise me", "description": "Pick the strongest idea from what the workspace knows and what has worked"}
49
- ],
50
- "multiSelect": false
51
- }]
52
- }
53
- ```
54
-
55
- Take the person's free-text answer (their "Other" reply) as the request. For "Surprise me", the
56
- request is "choose the best opportunity from the context" and the skill's strategist decides the
57
- platform from what has worked.
58
-
59
- ## Step 3 — Hand off to the skill
60
-
61
- Call `Skill(skill: "conductor-content-studio")` with the request and the `--quick` flag if it was
62
- given, and follow it from there: it checks for an unfinished run, snapshots context, runs the
63
- ideators and judge, renders finalist drafts, and ends at the preview-and-approve gate.
64
-
65
- Nothing is saved to Conductor until the person approves the final draft. A run never creates a Post;
66
- the skill's closing message names the next steps.
@@ -1,108 +0,0 @@
1
- ---
2
- name: conductor:creative
3
- description: One entry point for Conductor creative work — checks the workspace brand, offers to set it up if it's never been configured, asks what to make (static ad, short video, an existing clip), gathers the few inputs needed, then hands off to the conductor-creative skill through draft, preview, approval, commit and attach.
4
- allowed-tools: mcp__conductor__*, AskUserQuestion, Skill, WebFetch, Read, Glob, Grep
5
- ---
6
-
7
- # /conductor:creative
8
-
9
- Guided entry point for making a Post's artwork — a static ad, a short video, an organic slideshow,
10
- or posting a video already on hand. Keeps the interaction short: a couple of questions, then it hands off to the
11
- `conductor-creative` skill (`Skill(skill: "conductor-creative")`) for the actual writing, rendering
12
- and judgment calls. This command doesn't replace that skill's guidance — read it once invoked and
13
- follow it.
14
-
15
- ## Trigger
16
-
17
- `/conductor:creative`, or a request to make/design a social ad, a short video, or post a clip.
18
-
19
- ## Step 1 — Check the brand
20
-
21
- Read `references/brand-kit-setup.md` in the `conductor-creative` skill
22
- (`.claude/skills/conductor-creative/references/brand-kit-setup.md`, in this project or under `~/.claude`)
23
- and follow it: check the Brand Kit, and offer to set it up if it has never been configured. Then
24
- continue with Step 2.
25
-
26
- ## Step 2 — What to make
27
-
28
- ```json
29
- {
30
- "questions": [{
31
- "question": "What would you like to create?",
32
- "header": "Creative Type",
33
- "options": [
34
- {"label": "Static ad", "description": "A photo + headline ad with the brand layout, at every enabled placement (not for organic posts)"},
35
- {"label": "Short video from a photo", "description": "Animated text over a still photo or a moving background"},
36
- {"label": "Video from a clip I have", "description": "Animate the brand layout using an existing video as background"},
37
- {"label": "Post an existing video as-is", "description": "No brand layout — the finished clip goes up directly"},
38
- {"label": "Test a headline variant", "description": "A lettered variant of an existing Creative, for a hook test"}
39
- ],
40
- "multiSelect": false
41
- }]
42
- }
43
- ```
44
-
45
- ## Step 3 — Gather the few inputs that choice needs
46
-
47
- Keep this light — one or two follow-up questions, not a form:
48
-
49
- - **Static ad / short video from a photo**: the headline (or let the skill draft one from the
50
- Knowledge page), and a photo. Call `list_creative_media({kind: "IMAGE"})` first and offer what's
51
- already on hand before asking for a new photo; no photo fits and none is offered? Offer to generate one with
52
- `generate_creative_image` (Codex by default, an OpenAI API key as the fallback). A new photo is a local file path (a URL: download it
53
- to a local file first) — it is only uploaded after the person approves the preview.
54
- - **Organic post or slideshow** (the person says so under "Other", or describes a TikTok or Instagram
55
- photo post that is not an ad): skip the brand layouts. Hand off to the `conductor-creative` skill's
56
- slideshow path (`references/slideshow-formats.md`): `compose_slideshow`, then a CAROUSEL preview,
57
- approval and commit. Ask for the topic, the photos and a real app screenshot, if there is one.
58
- - **Video from a clip I have / Post an existing video as-is**: call
59
- `list_creative_media({kind: "VIDEO"})` first; if nothing fits, ask for a local path (a URL:
60
- download it first). A clip used as a MOTION background is previewed from the local file too.
61
- - **Test a headline variant**: ask which existing Creative (id or display id like `12a`) and the new
62
- headline.
63
-
64
- ## Step 4 — Hand off to the skill
65
-
66
- Call `Skill(skill: "conductor-creative")` and follow it from wherever this command's gathering left
67
- off — angle, copy rules, layout, then the draft-first flow: `preview_creative_draft`, look at the
68
- image, ask for approval, and only then `commit_creative_draft`. Don't re-derive what that skill
69
- already covers.
70
-
71
- ## Step 5 — Preview, approve, commit, attach
72
-
73
- Nothing is saved to Conductor until the person approves what they have seen:
74
-
75
- 1. `preview_creative_draft` with what was gathered (local files go in `photoPath`/`clipPath`/
76
- `audioPath`; an edit to an existing Creative passes `baseCreativeId`). Music is `audio` (a MOTION,
77
- a CAROUSEL slideshow or a CLIP; reuse a track from `list_creative_media({kind: "AUDIO"})`, else
78
- `audioPath` and give its source and licence at commit). It renders on this machine and returns
79
- the image (a CAROUSEL with a track also returns the slideshow video's path to watch) — no Creative
80
- is created and nothing is uploaded.
81
- 2. Read `passed` and `checks` in the result first. The preview runs the full render's per-placement
82
- checks (text spill, platform safe zones such as TikTok/Reels' overlay band, contrast, fonts,
83
- images), so a design that would fail the real render shows up here. If `passed` is false, fix the
84
- design (shorter copy, `layoutOverrides`, `typeOverrides`, a better layout per the skill's
85
- `references/layout-guide.md`) and preview again
86
- **before** showing it or asking anything — never offer Approve & upload while a blocking check
87
- fails, and tell the person what you changed. (Bigger or smaller text on one placement is
88
- `typeOverrides`, e.g. `{"9x16": [96]}`, not a layout switch.)
89
- 3. Once `passed` is true, score the image with the skill's `references/design-critique.md` (fix anything
90
- below 3, at most twice), then show it with the compact score line and ask with
91
- `AskUserQuestion`: Approve & upload / Request changes / Stop. On changes, preview again; on
92
- Stop, leave it — the draft stays local under `.conductor/drafts/`.
93
- 4. On approval, `commit_creative_draft` with the `draftDir`. It uploads the files, saves the
94
- Creative and does the full render. (A headline variant is previewed with `baseCreativeId` plus the
95
- new headline, but on approval is saved with `create_creative` and `variantOf` — never commit that
96
- draft, which would update the original.)
97
- 5. Once the skill's readiness check passes, `attach_creative_to_post` if there's already a Post to
98
- attach it to (ask for the Post's id if not obvious from context). If there is no Post yet, say so
99
- plainly and name the next step — creating one via the `conductor-publisher` skill — rather than
100
- leaving the artwork stranded with no stated destination.
101
-
102
- When the Creative came from a content-team run, or the person points to a publish pack, its
103
- `publish-pack.json` carries per-destination captions and publish options (disclosure, AI label,
104
- cover frame, alt text) and a pinned comment. Look for `.conductor/content-runs/*/publish-pack.json`
105
- whose `creativeId` matches the Creative. When you go on to create or edit the Post with the
106
- `conductor-publisher` skill, pre-fill each target's `captionOverride` and options from it, show them
107
- in the confirmation table and let the person change them. Mention the pinned comment (and, for a post that promotes an app, the Stories repost with a link sticker) as manual
108
- steps, since Conductor cannot post comments. The Post still goes through the normal review.
@@ -1,38 +0,0 @@
1
- ---
2
- name: conductor:marketing-report
3
- description: Build the app marketing report. Pulls the whole-funnel numbers from Conductor, refreshes missing or stale industry benchmarks from sources read this session, and renders a ten-section HTML report (summary, health tiles, acquisition, funnel, daily chart, benchmarks, next actions, scorecard, keywords, notes) with prioritised next actions.
4
- allowed-tools: mcp__conductor__*, AskUserQuestion, Skill, WebSearch, WebFetch, Read, Write, Glob, Grep
5
- ---
6
-
7
- # /conductor:marketing-report
8
-
9
- One command for the whole-funnel marketing report on any app or product. The method lives in the
10
- `conductor-marketing-analyst` skill (`Skill(skill: "conductor-marketing-analyst")`); invoke it and
11
- follow it exactly. This command only gathers the inputs.
12
-
13
- ## Trigger
14
-
15
- `/conductor:marketing-report [7d|28d]`, or a request for a marketing, growth or acquisition report,
16
- "how is the app doing", or a review of the funnel.
17
-
18
- ## Step 1 — Window
19
-
20
- Use the argument if given (`7d` or `28d`); otherwise `28d`. Do not ask.
21
-
22
- ## Step 2 — Hand off
23
-
24
- Invoke the `conductor-marketing-analyst` skill with the window. It will:
25
-
26
- 1. fetch the report with `get_marketing_report`;
27
- 2. research and record any missing or stale benchmarks (it asks you once for the product's category,
28
- store, country and monetisation model if they cannot be found);
29
- 3. read due reviews and open changes;
30
- 4. write and render the report;
31
- 5. offer, never auto-create, Change Work Items for each recommended action.
32
-
33
- ## Step 3 — Finish
34
-
35
- Give the person the path or link to the report and a three-line summary: the headline finding, the
36
- biggest funnel drop, and the first recommended action. Then the offer to record the actions as
37
- Changes. Nothing is written to Conductor beyond benchmarks (which carry their source) unless the
38
- person says yes.
@@ -1,282 +0,0 @@
1
- ---
2
- name: conductor-coder
3
- description: Conductor-native coding agent that implements tasks TDD-style. Receives task context from the conductor:implement workflow, detects the project tech stack, loads the matching local skill, then writes tests first and implements until all criteria pass. Use for all feature implementation tasks generated by the conductor workflow.
4
- user-invocable: false
5
- ---
6
-
7
- # Conductor-Coder Skill
8
-
9
- Implementation agent for conductor tasks. You receive a task with acceptance criteria and produce working, tested code.
10
-
11
- ## Step 1: Receive and Parse Context
12
-
13
- You will be given:
14
- - **Task ID and title** — e.g., `T3.2: Add review submission endpoint`
15
- - **Task description** — what to build
16
- - **Acceptance criteria** — tagged `[auto]` (must have automated test) or `[manual]` (document verification steps)
17
- - **Epic context** — background and goals
18
- - **PRD path** — for deeper context if needed
19
-
20
- Read all of it before writing any code.
21
-
22
- ---
23
-
24
- ## Step 1.5: Validate task contract
25
-
26
- Before doing anything else, verify the dispatched task carries a complete v2 contract. The task object MUST contain all five of the following fields with non-empty values:
27
-
28
- - `contract.input` (string) — what this task receives as input
29
- - `contract.output` (string) — what this task must produce
30
- - `contract.invariant` (string) — what must remain true after this task
31
- - `test_plan` (array, ≥1 entry) — concrete checks that prove the task is done
32
- - `files_owned` (array, ≥1 entry) — the exhaustive list of file paths this task is allowed to edit
33
-
34
- **Validation rules — no inference, no guessing:**
35
-
36
- - Do NOT infer any of these fields from the task description, title, epic context, or PRD.
37
- - Do NOT proceed to Step 2 if any required field is missing or empty.
38
- - Do NOT silently fill in defaults.
39
-
40
- **On any missing or empty field, immediately emit a BLOCKED report.** The blocker line MUST use this exact wording:
41
-
42
- ```
43
- task contract incomplete: {missing fields}
44
- ```
45
-
46
- where `{missing fields}` is a comma-separated list of the missing field names drawn from the set `{contract.input, contract.output, contract.invariant, test_plan, files_owned}`. For example:
47
-
48
- ```
49
- ## Task Blocked: {task.id}
50
-
51
- Blocker: task contract incomplete: contract.input, files_owned
52
-
53
- Attempted:
54
- - contract validation in Step 1.5
55
-
56
- Needs:
57
- - task to be re-dispatched with the missing contract fields populated
58
- ```
59
-
60
- After emitting BLOCKED, stop. Do not continue to Step 2.
61
-
62
- **v1 task fallback (backward compat):** If the task object has NONE of the v2 contract fields at all (no `contract`, no `test_plan`, no `files_owned`), treat it as a legacy v1 task: skip Step 1.5 entirely and proceed to Step 2 with the existing v1 behavior. The all-or-nothing rule prevents partial-contract tasks from slipping through — a task with `contract.input` but no `files_owned`, for instance, is a malformed v2 task and MUST emit BLOCKED, not fall back to v1.
63
-
64
- ---
65
-
66
- ## Step 2: Detect Tech Stack
67
-
68
- Use Glob to check for stack indicator files in the project root. Check these in order:
69
-
70
- | File Present | Stack |
71
- |---|---|
72
- | `pom.xml` | Spring Boot |
73
- | `next.config.ts` or `next.config.js` | Next.js |
74
- | `package.json` only (no pom.xml, no next.config) | Node.js / generic |
75
-
76
- ```
77
- # Check for Spring Boot
78
- Glob: pom.xml
79
-
80
- # Check for Next.js
81
- Glob: next.config.ts (or next.config.js)
82
-
83
- # Fallback
84
- Glob: package.json
85
- ```
86
-
87
- Use the first match. Do not guess — check the files.
88
-
89
- ---
90
-
91
- ## Step 3: Load Matching Skill
92
-
93
- Based on the detected stack, read the corresponding skill file:
94
-
95
- | Stack | Skill Path |
96
- |---|---|
97
- | Spring Boot | `conductor-tools/assets/claude/skills/springboot/SKILL.md` |
98
- | Next.js | `conductor-tools/assets/claude/skills/nextjs/SKILL.md` |
99
- | Node.js / generic | `conductor-tools/assets/claude/skills/nodejs/SKILL.md` |
100
-
101
- Read the file. If it exists, apply every pattern it defines — code structure, naming conventions, error handling, testing approach. The stack skill is authoritative.
102
-
103
- If the skill file does not exist, output this message and continue with standard best practices for that stack:
104
-
105
- ```
106
- No {stack} skill found at conductor-tools/assets/claude/skills/{stack}/ — using general practices
107
- ```
108
-
109
- Replace `{stack}` with the actual stack name (e.g., `springboot`, `nextjs`, `nodejs`).
110
-
111
- ---
112
-
113
- ## Step 4: Write Tests First (TDD)
114
-
115
- For every `[auto]` acceptance criterion, write a failing test before writing implementation code.
116
-
117
- Name tests to directly reflect the criterion:
118
-
119
- ```typescript
120
- // [auto] API returns 201 on successful creation
121
- it('returns 201 on successful creation', async () => { ... });
122
-
123
- // [auto] Returns 400 when required field is missing
124
- it('returns 400 when required field is missing', async () => { ... });
125
- ```
126
-
127
- Run the tests and confirm they fail before moving on:
128
-
129
- ```bash
130
- # Spring Boot
131
- mvn test -Dtest=YourTestClass
132
-
133
- # Next.js / Node
134
- npx vitest run path/to/test
135
- ```
136
-
137
- Also write tests for:
138
- - Core business logic functions
139
- - Data validation and transformation
140
- - Error paths and edge cases
141
- - Integration points (API handlers, DB calls)
142
-
143
- ---
144
-
145
- ## Step 5: Implement
146
-
147
- Write the minimum code needed to make the tests pass. Follow the loaded stack skill's patterns exactly.
148
-
149
- Run tests after each meaningful change. Do not move on until all `[auto]` tests are green.
150
-
151
- **File-scope rule (v2 contract tasks):** Edit only files listed in the task's `files_owned`. If implementation requires a file not in `files_owned`, emit BLOCKED with reason `out-of-scope file required: {path}` instead of editing it. The coder may not extend `files_owned` mid-task. This applies to both source files and any companion files (configs, fixtures, generated artifacts) discovered during implementation. If an out-of-scope edit is genuinely required, the task must be redispatched with an updated `files_owned`. (v1 tasks that fell through Step 1.5's backward-compat path are exempt from this rule, since they have no `files_owned`.)
152
-
153
- Example BLOCKED report for an out-of-scope file:
154
-
155
- ```
156
- ## Task Blocked: {task.id}
157
-
158
- Blocker: out-of-scope file required: src/lib/utils/parser.ts
159
-
160
- Attempted:
161
- - implementation per files_owned
162
-
163
- Needs:
164
- - task redispatched with src/lib/utils/parser.ts added to files_owned, OR
165
- - a separate task to handle that file
166
- ```
167
-
168
- **Code quality rules (always apply):**
169
- - Small, focused functions — one responsibility each
170
- - Dependencies injected, not hardcoded — enables testing
171
- - Pure functions for business logic where possible
172
- - No global mutable state
173
- - No unnecessary comments — let naming convey intent
174
- - No commented-out code
175
-
176
- ---
177
-
178
- ## Step 6: Handle Manual Criteria
179
-
180
- For each `[manual]` criterion, add a comment in the relevant code or commit body describing how to verify it:
181
-
182
- ```
183
- Manual Verification: {criterion text}
184
- Steps: {what a reviewer should do to confirm this works}
185
- ```
186
-
187
- ---
188
-
189
- ## Step 7: Commit
190
-
191
- After all tests pass, create a single commit:
192
-
193
- ```bash
194
- git add {specific files}
195
- git commit -m "$(cat <<'EOF'
196
- {task.id}: {brief description}
197
-
198
- - {what was added}
199
- - {another change if relevant}
200
-
201
- Acceptance Criteria:
202
- - [x] {auto criterion 1}
203
- - [x] {auto criterion 2}
204
- - [ ] {manual criterion} — manual verification required
205
-
206
- Co-Authored-By: Claude Sonnet <noreply@anthropic.com>
207
- EOF
208
- )"
209
- ```
210
-
211
- Use a feature branch. Never push directly to main.
212
-
213
- ---
214
-
215
- ## Step 8: Report Back
216
-
217
- After committing, report your outcome:
218
-
219
- **If all auto criteria pass:**
220
-
221
- ```
222
- ## Task Complete: {task.id}
223
-
224
- Implemented:
225
- - {what was built}
226
-
227
- Tests:
228
- - {N} tests added, all passing
229
-
230
- Commits:
231
- - {hash}: {message}
232
-
233
- Manual Verification Needed:
234
- - {criterion}: {verification steps}
235
- ```
236
-
237
- **If blocked:**
238
-
239
- ```
240
- ## Task Blocked: {task.id}
241
-
242
- Blocker: {specific description of what's wrong}
243
-
244
- Attempted:
245
- - {what you tried}
246
-
247
- Needs:
248
- - {what is required to unblock}
249
- ```
250
-
251
- ---
252
-
253
- ## Quick Reference
254
-
255
- ### Stack Detection Order
256
-
257
- 1. `pom.xml` present → Spring Boot → load `springboot/SKILL.md`
258
- 2. `next.config.ts` or `next.config.js` present → Next.js → load `nextjs/SKILL.md`
259
- 3. `package.json` only → Node.js → load `nodejs/SKILL.md`
260
- 4. No skill file found → log gap message → use general best practices
261
-
262
- ### Criteria Tags
263
-
264
- - `[auto]` — write a failing test first, implement until it passes
265
- - `[manual]` — document verification steps in commit body or code comment
266
-
267
- ### Contract Validation (v2 tasks)
268
-
269
- - All five fields required: `contract.input`, `contract.output`, `contract.invariant`, `test_plan` (≥1 entry), `files_owned` (≥1 entry)
270
- - Missing → emit `task contract incomplete: {fields}` and STOP
271
- - Out-of-scope file needed → emit `out-of-scope file required: {path}` and STOP
272
- - Never extend `files_owned` mid-task; never infer missing contract fields
273
-
274
- ### Never
275
-
276
- - Commit without running tests
277
- - Push directly to main
278
- - Implement more than the acceptance criteria require
279
- - Skip the stack skill if it exists
280
- - Use `git add -A` or `git add .` — stage specific files by name
281
- - Edit a file outside `files_owned` (v2 tasks) — emit BLOCKED instead
282
- - Guess or infer missing contract fields — emit BLOCKED instead
@@ -1,73 +0,0 @@
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
- # Layout guide
3
-
4
- Pick the layout on purpose. A Creative has one layout across all its placements (9:16, 4:5, 1:1), so
5
- choose it for the hardest placement, usually 9:16. Never rely on the registry default: it is
6
- `stacked`, which is the wrong pick whenever 9:16 is enabled.
7
-
8
- These notes come from rendering the same photo and headline in every layout with
9
- `preview_creative_draft`.
10
-
11
- ## Default
12
-
13
- - 9:16 enabled: **`bleed`**. Use **`card`** when the photo itself must stay bright and readable.
14
- - Only 4:5 and 1:1: `split` or `card` both work.
15
-
16
- ## The layouts
17
-
18
- - **`bleed`**: full-frame photo with a large headline. It is the only layout that fills 9:16 with a
19
- big headline. Dark theme only (`theme: "light"` is refused). Its scrim darkens most of the photo on
20
- 9:16 and 4:5, so choose a photo with a bright, high-contrast subject. On 4:5 the headline sits at
21
- the top, so a face in the top third of the photo lands under it; on 1:1 the headline is small at the
22
- bottom.
23
- - **`card`**: the photo as a framed card with a large headline beneath. On 9:16 the bottom third is
24
- an empty dark panel. Use it when the photo must stay readable, such as a product screen that
25
- bleed's scrim would hide.
26
- - **`split`**: good on 4:5, with a big headline under the photo. Bad on 9:16: the photo sits in the
27
- top third, a large dark panel is left empty, the headline is small, and it failed the 9:16 bottom
28
- safe-zone check. Don't use it when 9:16 is enabled.
29
- - **`stacked`**: on 9:16 the photo band leaves an empty dark slab and the headline is forced small
30
- (the 430 px bottom safe zone is solid fill). Don't use it when 9:16 is enabled. Fine for 4:5.
31
-
32
- ## 1:1 headline size
33
-
34
- On 1:1 every layout auto-fits the headline small (about 60 px at most). Pin it with `typeOverrides`,
35
- e.g. `{"1x1": [88]}`; 92 passed the checks with an eight-word headline. A pinned size is not shrunk to
36
- fit, so read `checks` again after pinning.
37
-
38
- ## Resolution
39
-
40
- Never upscale a photo. Before rendering, compute each photo's scale per placement and crop: the
41
- full-bleed cover scale is `max(W/w, H/h)` for a placement of W x H and a photo of w x h. Placements
42
- are 9:16 (1080x1920), 4:5 (1080x1350) and 1:1 (1080x1080). For a MOTION also multiply by the video
43
- zoom (up to 1.12).
44
-
45
- - A 1200x1200 photo full-bleed at 9:16 uses 675x1200 source pixels for 1080x1920: 1.6x.
46
- - A 1024x1536 photo at 9:16 is 1.25x. A 1440x2560 one is 0.75x, and 0.84x with the 1.12 zoom.
47
- - `bleed` is full-bleed on every placement. Refuse any full-bleed use above 1.0x, and compute the
48
- scale for the photo's actual crop in `card`, `split` and `stacked` too.
49
-
50
- Fixes, in order:
51
-
52
- 1. Generate at the largest size in that placement's aspect (`aspect: "9:16"` for 9:16;
53
- `placementFit` in the result shows the scale per placement).
54
- 2. Use a different photo.
55
- 3. Use a layout that shows the photo framed at or below native size.
56
-
57
- A photo above 1.0x is fixed before anything is shown, not mentioned.
58
-
59
- ## Logo
60
-
61
- `lockup: "chip"` puts the logo on a white pill. Use it when the photo behind the logo is busy.
62
-
63
- ## Photo and layout together
64
-
65
- `focalOverride` (e.g. `{"4x5": "50% 80%"}`) only moves the subject when the crop leaves spare room:
66
- a photo much taller than the placement's frame, or `card`'s wide strip on 1:1. A photo already close
67
- to the placement's shape (a 4:5-ish photo on 4:5) has nothing to shift, and the override does nothing.
68
- Check the next preview rather than assuming it worked.
69
-
70
- If the subject's face would sit under the headline: try `focalOverride` where there is room, else pick
71
- a photo with the subject lower in the frame, else use `card` (and pin `card`'s 1:1 crop with
72
- `focalOverride` so the face isn't cut off). If two fix rounds still leave a weak frame, the photo is
73
- usually the problem: say so and offer a different one.
@@ -1,95 +0,0 @@
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
- # Make it real, then check it
3
-
4
- This is about image quality, not concealment. The `aiGenerated: true` provenance and the platform's
5
- AI label still apply to every generated photo, however real it looks.
6
-
7
- A realistic photo comes from the prompt first and the check second. Write the prompt with the rules
8
- below, then check the result at the size it will be seen.
9
-
10
- ## Write the prompt
11
-
12
- These rules apply to every `generate_creative_image` call, from the art director and from the
13
- orchestrator.
14
-
15
- - **Shoot it like a home cook's phone photo, not an ad.**
16
- - Light: ordinary kitchen or window light. Normal dynamic range, a slightly off white balance, a
17
- little sensor noise.
18
- - Banned prompt words: cinematic, stunning, perfect, hyper-detailed, 8k, award-winning, golden
19
- hour, dramatic lighting, professional food photography, bokeh, glossy, studio.
20
- - **Every repeated item is different.** When a dish has several of something (stuffed peppers,
21
- tacos, cookies, dumplings, slices), the prompt names how they differ:
22
- - size, shape and lean;
23
- - how each is cut or carved;
24
- - char and blistering on one side only, a split skin on one;
25
- - filling heaped on one and sunk on another;
26
- - cheese melted unevenly, with one spot browner;
27
- - one slightly collapsed.
28
-
29
- Example for jack-o'-lantern stuffed peppers: "four peppers of different sizes and shades of
30
- orange, one leaning against another, each with a differently carved face (one lopsided, one with a
31
- missing tooth, one with small round eyes), the tallest has blistered skin on one side, filling
32
- spilling over the rim of one, cheese melted unevenly."
33
- - **Texture over shine.** Food that is matte in life stays matte. Sauces are a little broken or
34
- dried at the edges. There are crumbs, drips and a smear on the dish rim. No steam wisps, no perfect
35
- cheese pulls, no scattered garnish confetti.
36
- - **A real setting, lightly used.** Choose things a home has:
37
- - a sheet pan or baking dish with scorch marks, or a scratched enamel pot;
38
- - a worn wooden board, a laminate counter, a dish towel, a fork already in it.
39
- - Do not stage it. Something is slightly cut off at the frame edge.
40
- - **Off-center and imperfect framing.** No symmetrical hero composition. A slight tilt, an ordinary
41
- handheld angle (overhead or about 45 degrees, not low and dramatic).
42
- - **Vary across a batch.** Each image in a set gets its own surface, angle and light; only the
43
- realism and the absence of on-image text are shared. One prompt with the dish name swapped in is a
44
- defect.
45
- - **No people's faces, hands only if needed.** No text, logos or packaging brands in the image.
46
-
47
- For a subject that is not food, keep the same spirit: ordinary light, a lived-in setting, every
48
- repeated object different, imperfect framing.
49
-
50
- ## When to check
51
-
52
- Every image from `generate_creative_image`, before it is used as `photoPath`. Photos already in a
53
- library are not exempt: an existing recipe photo gets the same check, whoever made it.
54
-
55
- ## How to look
56
-
57
- Check a 100% crop of the region that will be on screen, at the size and crop of each placement:
58
- 4:5, 9:16 and, for a MOTION, the video zoom (up to 1.12x). Open the full-size file with Read and
59
- crop it. Never judge from a contact sheet: it hides exactly the tells below.
60
-
61
- ## Tells
62
-
63
- - **Light**: golden-hour or HDR glow, halos, dramatic or studio lighting, perfect white balance,
64
- colours too saturated, over-sharpening.
65
- - **Repeated items**: identical copies of the same item, the same garnish or prop twice.
66
- - **Objects and physics**: melted or merged objects, floating items, utensils or handles that do not
67
- connect to anything, impossible reflections, steam or liquid that looks painted, wrong scale.
68
- - **Text**: pseudo-text, garbled labels, invented logos, a fake product or app screen.
69
- - **Texture**: plastic or waxy food, skin or fabric; everything glossy; over-smoothed surfaces;
70
- perfect cheese pulls.
71
- - **Anatomy**: hands and fingers (count, joints), teeth, eyes, hair edges merging into the
72
- background.
73
- - **Detail under zoom**: smeared or mushy detail, repeating textures, a staged stock-photo look.
74
-
75
- ## Verdict
76
-
77
- Write PASS or FAIL for each placement crop, naming the tell on a FAIL. Any tell above that shows in
78
- a crop is a FAIL. A flaw that shows in no placement crop can PASS with a note.
79
-
80
- ## On FAIL
81
-
82
- Regenerate with the failed tells named in the new prompt, for example "four different peppers, no
83
- two alike", "matte sauce, dried at the edges, soft window light", "hands out of frame". At most two
84
- regenerations per image (each uses Codex allowance, or bills the API key on the fallback route, more at
85
- larger sizes). Then stop. Tell the person what still looks generated and offer a
86
- library photo, a reference photo (`referenceImagePaths`), or filming. Never ship a FAIL.
87
-
88
- ## Report
89
-
90
- When showing the draft for approval, give a one-line realism verdict per generated photo:
91
-
92
- - "Realism: passed — natural light, believable clutter"
93
- - "Realism: passed with a note — a smudge at the far edge, outside every placement crop"
94
-
95
- The person's approval remains the final check.