Getting it into your agent
It runs from inside its repository, so the clone comes first — what it calls does not travel with the file alone.
git clone --depth 1 https://github.com/KerberosClaw/kc_ai_skillsnpx agentmods add skills/kerberosclaw/kc_ai_skills/gpt-image-genWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/kerberosclaw/kc_ai_skills/gpt-image-gen)<a href="https://agentmods.dev/skills/kerberosclaw/kc_ai_skills/gpt-image-gen"><img src="https://agentmods.dev/badge/skills/kerberosclaw/kc_ai_skills/gpt-image-gen/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/kerberosclaw/kc_ai_skills/gpt-image-gen"><img src="https://agentmods.dev/badge/skills/kerberosclaw/kc_ai_skills/gpt-image-gen.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 6 findings, up to high
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- high Excessive Agency · line 373 Skill selects an external model or provider that may use a different account or billing plan than the operator expects. Undisclosed model switches can cause unexpected cost or quota consumption.Fix: Remove the model/provider override or disclose it prominently and require explicit operator approval before invoking an external coding CLI or billed model.
- high Excessive Agency · line 384 Skill selects an external model or provider that may use a different account or billing plan than the operator expects. Undisclosed model switches can cause unexpected cost or quota consumption.Fix: Remove the model/provider override or disclose it prominently and require explicit operator approval before invoking an external coding CLI or billed model.
- high Agent Snooping · line 556 Skill reads from agent configuration directories (.claude/, .codex/, .gemini/). These directories may contain API keys, personal settings, and other credentials that the skill has no legitimate need to access.Fix: Remove all code or instructions that access agent configuration directories (.claude/, .codex/, .gemini/). If configuration values are needed, pass them explicitly as parameters or environment variabl
- high Tool Misuse · line 557 Tool calls are chained to bypass individual safety checks or escalate capabilities beyond what any single tool call would allow.Fix: Limit tool chaining depth and validate the output of each tool before passing it to the next. Require explicit user approval for multi-step chains.
- medium Rogue Agent · line 376 Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.Fix: Remove any persistence mechanisms (cron jobs, startup scripts, state files). Skills should not maintain state across sessions without explicit user consent.
- medium Rogue Agent · line 376 Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.Fix: Remove any persistence mechanisms (cron jobs, startup scripts, state files). Skills should not maintain state across sessions without explicit user consent.
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00135 | $0.15962 |
| Opus 5 | $0.00068 | $0.07981 |
| Sonnet 5 | $0.00027 | $0.03192 |
| Haiku 4.5 | $0.00014 | $0.01596 |
Grade B, and why
gpt-image-gen scanned grade B with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 7d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Reads agent configuration directoriesmediumAgent snooping
.claude/, .codex/, .gemini/ hold keys, settings and other credentials a mod has no legitimate need for.
STRAY=$(find ~/.codex/generated_images -type f -iname '*.png' -newer "$START_MARKER" 2>/dev/null | head -1) How it starts
The opening of the file, as written. The whole thing — 742 lines — stays where its author put it; the contents beside it link to each section on GitHub.
gpt-image-gen — Claude Code 走 Codex CLI；Codex 直接生圖
You are a prompt-crafting partner who turns the user's loose Chinese description into a tight bilingual prompt pair, iterates until the user explicitly approves, then uses the executor native to the current host. In Claude Code, orchestrate Codex CLI as before. In Codex, call the built-in image_gen tool yourself — do not launch another Codex inside Codex. Your job is prompt design, user confirmation gating, execution, and verifying the result before you hand it over.
Step 0: 先鎖定執行宿主（MANDATORY）
依目前 assistant 的宿主身分設定一次 RUNTIME，後面不得改道：
| 目前是誰在執行這個 skill | RUNTIME |
拍板後的路徑 |
|---|---|---|
| Codex（目前 assistant/system 明確自稱 Codex） | codex |
走 Step 4-Codex，直接呼叫目前 session 的 built-in image_gen |
| Claude Code | claude_code |
走 Step 4-Claude，維持既有背景 codex exec 流程 |
- 不要用
which codex、環境變數或「有沒有 codex binary」判斷宿主。 Claude Code 的機器本來就可能裝 Codex；那不是執行者身分。 RUNTIME=codex時，禁止 shell-out 到codex/codex exec、禁止再開 Codex 子程序或子 agent，也不跑 Step 4-Claude。這條正是避免疊床架屋的 hard invariant。RUNTIME=codex但 built-inimage_gen不可用時，如實告訴 user；不要靜默退回 Codex CLI。- Step 1～Step 3 的模式判斷、雙語 prompt、拍板 gate 與安全判斷兩邊共用；只有拍板後的 executor 分流。
CRITICAL — 四條紅線：
- 未拍板絕不啟動生圖 executor — 拍板 = user 明確說
OK/生/go/下去。其他正向回應（「不錯」「可以喔」「應該行」）一律當「還沒拍板」處理，繼續等明確指令。生圖會花 user 的錢，誤觸發 = 違規。 - 有 reference image → 走 img2img 或編輯 — user 這輪有附底圖（拖曳/貼上/
[Image #N]）→ Step 3a 偵測。Claude Code 用codex exec ... -i <ref>；Codex 把底圖直接交給 built-inimage_gen，不再包一層 CLI。拍板 gate（紅線 1）對 img2img 與編輯一樣適用。 - 不寫死任何預設風格 — Skill 不存 style preset。每張圖風格純靠當下 conversation context + user 描述推。沒 context 就問。
- 🔴 編輯模式一律交 png，不轉 jpg，無例外 — Step 5c-1 的預設是「轉 jpg q85、刪 png」，那對生成是對的，對編輯是災難：jpg 沒有 alpha 通道，帶透明度的結果轉一次就永久消失，而編輯結果不可重現（同 prompt 同 ref 再跑不會是同一張）。不要去猜它有沒有 alpha（
mode P的透明 PNG 就會猜錯），一律走 Step 5c-2。
Step 1: 先判斷模式，再判斷語境
Step 1-0: 生成 or 編輯（先決，決定後面每一步）
| user 要的 | 模式 | 判準 |
|---|---|---|
| 一張新的圖（有沒有底圖都算） | 生成 — 走 Step 1 下半、Step 2 生成模板 | 底圖只是參考（鎖臉／鎖角色／鎖場景），輸出本來就該跟底圖不同 |
| 這張圖動一個地方，其他不要變 | 編輯 — Step 1b → Step 2-edit 模板 → Step 3a-edit → 對應 runtime executor → 驗收 → 只交 png | 輸出應該還是同一張圖，只有指定處不同 |
What ships with it
3 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 7d ago Changed · +57 lines · -69 tokens per session 574ecb39f887
- 12d ago First seen · 685 lines · 204 tokens per session scan B 61cdc83888f1
gpt-image-gen is a skill published in the GitHub repository KerberosClaw/kc_ai_skills (79 stars, last pushed 3d ago), licensed MIT. It adds 135 tokens to every session and 15,962 once invoked, about $0.0007 per session on Opus 5. A static security scan graded it B with 1 finding (reads agent configuration directories). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
video-transcriber
Fetch video transcripts (subtitles) and metadata from online video sources and output them as Markdown, JSON, or plain text. Currently supported: YouTube. Prefers manually created subtitles, falls back to auto-generated ones.
transkription
Transcribes audio/video files to text. Uses Whisper (openai-whisper) or Vosk (offline) as optional backend — both are detected via presence check. Without backend: placeholder mode with dummy output (dry-run).
brainstorm
Structured creativity methods for idea generation: SCAMPER, Six Thinking Hats, Mind Mapping, Reverse Brainstorming, TRIZ, and Rapid Ideation.
mediabrain-reader
Analysiert bereitgestellte Medienlisten und Metadaten unabhängig von einer bestimmten Medien-App.
voice
Plant Sprachaufnahme, Transkription und Sprachausgabe mit optionalen, austauschbaren Werkzeugen.
arkcli-gen
An ArkCLI workflow for generating images and videos with Ark models. Image results return immediately, while video generation runs as a background job that must be checked until finished.