Create Image Gpt Image Fal
Python CLI for gpt-image-1/gpt-image-2 on fal.ai, routed through the GooseWorks proxy
Test report
- Verdict
- Works with setup
- Score
- Tested
- Jul 21, 2026
- Environment
- Claude Code 2.x (agent harness)
- Upstream re-checked
- Aug 10, 2026 · 7e2e791
Located the skill via `gh api .../git/trees/HEAD?recursive=1` (raw GitHub API was rate-limited) at skills/ads/capabilities/create-image-gpt-image-fal/SKILL.md; all five referenced files fetched HTTP 200 (scripts/generate.py 7060B, scripts/media_proxy.py 10527B, scripts/fal_helpers.py 5496B, skill.meta.json, tests/smoke-test.md). I could not generate an image: ~/.gooseworks/credentials.json does not exist and the script is proxy-routed through a paid GooseWorks fal-proxy, so I ran only the offline argument paths — local `--ref-image` correctly exits with "ERROR: ref images must be PUBLIC URLs", `--aspect-ratio 4:5` on gpt-image-1 errors with the supported list, and `--image-size 1728x2304` on gpt-image-1 prints the documented warning then falls back to 1024x1536 before crashing with a raw FileNotFoundError traceback on the missing credentials file (not the friendly "run gooseworks login" the Preflight implies). Two doc defects verified by execution: the failure-modes table claims "The script rounds to /16 and caps at 3840", but `--model gpt-image-2 --image-size 4000x4000` exits with "ERROR: image size exceeds 3840px limit"; and bundled tests/smoke-test.md tells you to confirm FAL_API_KEY in .env and calls a stale path (skills/atoms/image-generation/...) that does not exist in the repo, directly contradicting SKILL.md's "Do not set FAL_API_KEY". Security scan: no curl|sh, no base64 blobs, no third-party exfiltration — but media_proxy.py sends the agent token as a URL query param (?token=&agent_id=) rather than an Authorization header, which leaks into any proxy/server logs. Trigger phrasings judged: SHOULD fire — "Generate a photoreal 9:16 anchor portrait for our ad character with gpt-image on fal" (load), "I need a 1728x2304 designed storyboard sheet, use gpt-image-2 with a custom output size" (load), "Here's the public URL of the approved anchor, produce the 3/4-left angle keyframe from it" (load, edit variant); SHOULD NOT fire — "Animate this keyframe into a 5-second clip" (correctly skip, video atom), "Overlay the headline 'Klarify - book in 60s' onto the finished creative" (I skip, but only because the body says to use PIL/ffmpeg for readable text — the frontmatter description's "designed sheets ... where precise layout and legible text matter" pulls the other way, so 4/5).
Scored on four weighted criteria — install, triggering, output vs. baseline, docs. How scoring works
- Installs cleanly 4/5
- Triggers reliably 4/5
- Output vs. baseline 2/10
- Docs & honesty 3/5
What Create Image Gpt Image Fal does
Wraps fal.ai's OpenAI gpt-image endpoints in a single Python CLI that generates one image and writes a sidecar meta.json with model, size, quality and cost estimate. It selects gpt-image-1 (three fixed sizes) or gpt-image-2 (custom sizes up to 3840px), and switches between the text-to-image and edit variants depending on whether public reference-image URLs are passed. Triggers when generating photoreal character anchors, scene keyframes, or designed storyboard sheets through fal.ai rather than a generic image request.
How to install Create Image Gpt Image Fal
git clone --depth 1 https://github.com/gooseworks-ai/goose-skills.git /tmp/create-image-gpt-image-fal-src
mkdir -p ~/.claude/skills
cp -R /tmp/create-image-gpt-image-fal-src/skills/ads/capabilities/create-image-gpt-image-fal ~/.claude/skills/create-image-gpt-image-fal
# Paid GooseWorks fal proxy — the skill cannot render an image without credentials:
# npx gooseworks login # writes ~/.gooseworks/credentials.json
# export GW_PROJECT_ID=<ad-project-id> # so spend attributes to the right project
# pip3 install requests # undeclared dependency
# Run: python3 ~/.claude/skills/create-image-gpt-image-fal/scripts/generate.py \
# --prompt "..." --output out.png --aspect-ratio 9:16 --quality medium
Skills live in ~/.claude/skills/ (global) or .claude/skills/
(per-project). Restart Claude Code after installing.
Commands — how to trigger Create Image Gpt Image Fal
-
/create-image-gpt-image-falPython CLI for gpt-image-1/gpt-image-2 on fal.ai, routed through the GooseWorks proxy
It also activates on plain-language prompts like these:
-
Generate a photoreal character anchor image -
Create a scene keyframe with a reference image -
Design a storyboard sheet with legible text
Frequently asked questions
- Is the Create Image Gpt Image Fal skill free?
- Yes. The skill itself is free from gooseworks-ai/goose-skills. SkillProof publishes the install command and an independent test verdict at no cost.
- Does Create Image Gpt Image Fal work with Claude Code?
- We tested it with Claude Code 2.x (agent harness) on Jul 21, 2026. Verdict: Works with setup. Located the skill via `gh api .../git/trees/HEAD?recursive=1` (raw GitHub API was rate-limited) at skills/ads/capabilities/create-image-gpt-image-fal/SKILL.md; all five referenced files fetched HTTP 200 (scripts/generate.py 7060B, scripts/media_proxy.py 10527B, scripts/fal_helpers.py 5496B, skill.meta.json, tests/smoke-test.md). I could not generate an image: ~/.gooseworks/credentials.json does not exist and the script is proxy-routed through a paid GooseWorks fal-proxy, so I ran only the offline argument paths — local `--ref-image` correctly exits with "ERROR: ref images must be PUBLIC URLs", `--aspect-ratio 4:5` on gpt-image-1 errors with the supported list, and `--image-size 1728x2304` on gpt-image-1 prints the documented warning then falls back to 1024x1536 before crashing with a raw FileNotFoundError traceback on the missing credentials file (not the friendly "run gooseworks login" the Preflight implies). Two doc defects verified by execution: the failure-modes table claims "The script rounds to /16 and caps at 3840", but `--model gpt-image-2 --image-size 4000x4000` exits with "ERROR: image size exceeds 3840px limit"; and bundled tests/smoke-test.md tells you to confirm FAL_API_KEY in .env and calls a stale path (skills/atoms/image-generation/...) that does not exist in the repo, directly contradicting SKILL.md's "Do not set FAL_API_KEY". Security scan: no curl|sh, no base64 blobs, no third-party exfiltration — but media_proxy.py sends the agent token as a URL query param (?token=&agent_id=) rather than an Authorization header, which leaks into any proxy/server logs. Trigger phrasings judged: SHOULD fire — "Generate a photoreal 9:16 anchor portrait for our ad character with gpt-image on fal" (load), "I need a 1728x2304 designed storyboard sheet, use gpt-image-2 with a custom output size" (load), "Here's the public URL of the approved anchor, produce the 3/4-left angle keyframe from it" (load, edit variant); SHOULD NOT fire — "Animate this keyframe into a 5-second clip" (correctly skip, video atom), "Overlay the headline 'Klarify - book in 60s' onto the finished creative" (I skip, but only because the body says to use PIL/ffmpeg for readable text — the frontmatter description's "designed sheets ... where precise layout and legible text matter" pulls the other way, so 4/5).
- What is the Create Image Gpt Image Fal SkillProof Score?
- 5.2/10 — installs cleanly 4/5, triggers reliably 4/5, output vs. baseline 2/10, docs & honesty 3/5.
- How do I install Create Image Gpt Image Fal?
- Copy the install command from this page, run it in your terminal, and restart Claude Code. Skills live in ~/.claude/skills/ (global) or .claude/skills/ inside a project.
- Can I use Create Image Gpt Image Fal with Cursor, Copilot, Gemini CLI, Codex or other AI tools?
- The SKILL.md format is native to Claude (Claude Code, Desktop, claude.ai). The instructions inside adapt to other assistants: Cursor rules, GitHub Copilot instructions, Windsurf rules, Custom GPTs, AGENTS.md for OpenAI Codex, and GEMINI.md for Google Gemini CLI — our conversion guides cover each, and the free converter on the tools page does the wrapping for you.