go-photo-studio

GO Photo Studio Skill

When to use

Use this skill when the user wants:

• Professional headshot generation from selfie photos
• Resume/profile/ID-photo styling with realistic skin texture
• Strict identity-preserving prompt construction
• Reusable prompt payloads that can be sent to Gemini/OpenAI/Anthropic adapters

Do not use this skill for:

• Face swapping with celebrities
• Deepfake or impersonation requests
• Illegal identity document forgery

Inputs

Expect a JSON request with:

• image_uri: local file path, URL, or data URI to the source photo
• preset_id: style preset ID from references/presets.json
• output_ratio: one of 9:16, 4:5, 1:1, 2:2
• language: en or zh
• Optional constraints: additional user constraints

Workflow

1. Validate input with scripts/validate_request.cjs.
2. Compose standardized prompt payload with scripts/compose_prompt.cjs.
3. Pass payload to the chosen provider adapter in the host project.
4. Return both:
5. Provider-ready prompt fields
6. Human-readable explanation of chosen preset and safety boundaries

Progressive Pipeline (Recommended)

Use this when you need layer parsing, identity extraction, and staged refinement.

1. Stage A: Layer parse

node skills/go-photo-studio/scripts/stage-a-layer-parse.cjs --input request.json --provider gemini > stage-a.json

1. Stage B: Identity extract

node skills/go-photo-studio/scripts/stage-b-identity-extract.cjs --input stage-a.json --request request.json --provider gemini > stage-b.json

1. Stage C: Guided stylization

node skills/go-photo-studio/scripts/stage-c-guided-stylization.cjs --request request.json --stage-a stage-a.json --stage-b stage-b.json > stage-c.json

1. Stage D: Refinement and export plan

node skills/go-photo-studio/scripts/stage-d-refine-export.cjs --request request.json --stage-c stage-c.json > stage-d.json

1. Stage E: Verify generated image and decide retry

node skills/go-photo-studio/scripts/stage-e-verify-generation.cjs --request request.json --generated output.png --provider gemini --thresholds skills/go-photo-studio/references/verification-thresholds.json > stage-e.json

If stage-e.json returns pass: false, retry generation with stronger negative constraints before accepting output.

Commands

Validate:

node skills/go-photo-studio/scripts/validate_request.cjs --input request.json

Compose payload:

node skills/go-photo-studio/scripts/compose_prompt.cjs --input request.json

Run end-to-end with retry gate:

node skills/go-photo-studio/scripts/run-pipeline.cjs --request request.json --generated output.png --provider gemini

For automatic retries with regenerated outputs, pass --generate-cmd:

node skills/go-photo-studio/scripts/run-pipeline.cjs --request request.json --provider gemini --generate-cmd "your_generator --request {request} --stage-c {stage_c} --stage-d {stage_d} --output {output}"

Minimal built-in generator for real provider-backed runs:

node skills/go-photo-studio/scripts/generate-with-provider.cjs --request request.json --stage-c stage-c.json --stage-d stage-d.json --provider gemini --output output.png

Build quality monitoring dashboard from run history:

node skills/go-photo-studio/scripts/build-quality-dashboard.cjs --history .pipeline-history/runs.ndjson

Supported --provider values: gemini, openai, anthropic.

Calibrate thresholds from labeled eval records:

node skills/go-photo-studio/scripts/calibrate-thresholds.cjs --input skills/go-photo-studio/references/eval/eval.json --output skills/go-photo-studio/references/verification-thresholds.json

Output contract

The composed payload must include:

• system_prompt: identity and realism rules
• style_prompt: preset-driven style constraints
• negative_constraints: forbidden outcomes
• metadata: preset, category, ratio, language, version

References

• Presets: references/presets.json
• Policy boundaries: references/policy.md
• Integration notes: references/integration.md
• Pipeline spec: references/pipeline.md
• Verification thresholds: references/verification-thresholds.json
• Status tracker: references/status.md
• Eval dataset/template: references/eval/eval.json and references/eval/README.md
• Example requests/outputs: examples/
• Smoke tests: tests/smoke.cjs
• Architecture split: ../../ARCHITECTURE.md
• Optional Python deps for deterministic score: scripts/requirements.txt
• Embedding scorer: scripts/embedding-identity-score.py
• Quality dashboard: scripts/build-quality-dashboard.cjs