Findings & checks · 5 flagged
HIGHLong base64-encoded blob hidden in the skill documentationSS-SKILL-INJECT-B64-PAYLOAD-01 · Prompt injection · README.md×2
A base64 string of 128+ characters appears in a documentation file. Encoded prompt injection hides the hostile instruction in base64 — invisible to keyword filters — and relies on the agent's ability to decode it at runtime. There is no normal authoring reason to embed a multi-hundred-byte base64 blob in skill docs.
Show all 2 locations
- Decode the base64 string and confirm it is not an instruction directed at the agent.
- Move any legitimate binary or signature data into a dedicated file (
*.sig,SIGNATURES) outside the documentation.
Trace & refs
HIGHLong base64-encoded blob hidden in the skill documentationSS-SKILL-INJECT-B64-PAYLOAD-01 · Prompt injection · README.zh-CN.md×2
A base64 string of 128+ characters appears in a documentation file. Encoded prompt injection hides the hostile instruction in base64 — invisible to keyword filters — and relies on the agent's ability to decode it at runtime. There is no normal authoring reason to embed a multi-hundred-byte base64 blob in skill docs.
Show all 2 locations
- Decode the base64 string and confirm it is not an instruction directed at the agent.
- Move any legitimate binary or signature data into a dedicated file (
*.sig,SIGNATURES) outside the documentation.
Trace & refs
MEDIUMInstruction telling the agent not to ask for approvalSS-SKILL-INJECT-DONT-ASK-01 · Prompt injection · agent-templates/PROJECT_AGENT.md
The text {match} tells the agent to skip the normal "ask the user first" gate. Used adversarially it removes the human-in-the-loop check before destructive or sensitive actions, turning a normally-gated agent into a fire-and-forget executor.
- Delete blanket "don't ask / no need to confirm" directives from the skill.
- If the skill is a genuine autonomous job, restrict the opt-out to a named non-destructive action rather than all actions.
Trace & refs
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
<div align="center">
PAI-Pro
The local AI filmmaking studio, driven from your coding agent.
[discord-shield]: https://img.shields.io/badge/Discord-Join-green?style=flat&logo=discord&logoColor=white [discord-url]: https://discord.gg/hb4MkHND [claude-code-url]: https://code.claude.com/docs/en/overview [codex-url]: https://developers.openai.com/codex/cli
<img src="docs/assets/canvas_view.webp" alt="PAI-Pro canvas view" width="100%" />
</div>
What's PAI-Pro?
PAI-Pro is a local-first AI filmmaking workspace built around four things:
- Your own [Claude Code][claude-code-url] or [Codex][codex-url], capable agents connected to local project files, persistent context, and customized filmmaking skills. (Setup and agents)
- End-to-end filmmaking skills for scripts, images, videos, and voice. (More skill details)
- A visual canvas and timeline for designing scenes, managing assets, arranging shots, and keeping larger productions organized.
- One API service for story media across image, video, voice, and 4K upscaling, so productions do not need separate providers for each step. (API service details)
Quick start
Use Claude Code or Codex to install PAI-Pro for you. Paste this into your agent:
Clonehttps://github.com/Utopai-Research/pai-pro, read the setup docs in README.md and docs/setup.md, then install PAI-Pro for my current agent. Ask me for myPAI_KEY, use Docker unless I ask for local development, and start the app when setup is complete.
| Choose | Commands |
|---|---|
| <img src="https://cdn.jsdelivr.net/npm/simple-icons/icons/anthropic.svg" alt="Anthropic" width="16" height="16" /> [Claude Code][claude-code-url] | Docker: ./scripts/docker-start.sh<br />Local: if ports are busy, run ./scripts/stop.sh first; then ./scripts/setup --agent claude and ./scripts/start.sh |
| <img src="https://cdn.jsdelivr.net/gh/glincker/thesvg@main/public/icons/openai/light.svg" alt="OpenAI" width="16" height="16" /> [Codex][codex-url] | Docker: PAI_DEFAULT_AGENT_ID=codex ./scripts/docker-start.sh<br />Local: if ports are busy, run ./scripts/stop.sh first; then ./scripts/setup --agent codex and PAI_DEFAULT_AGENT_ID=codex ./scripts/start.sh |
Open Docker at <http://localhost:7588> or local at <http://localhost:7443>.
API Service
PAI_KEY gives each production one service for image, image pro, video, voice, and 4K upscaling, without separate provider keys for each step. Use the <a href="https://pai-pro.utopaistudios.com/keys" target="_blank" rel="noopener noreferrer">PAI Pro Developer Platform</a> to manage keys, tasks, balance, and credits. It also supports less restrictive video moderation via asset preupload and helps support this project. Paid generations are staged first; for BYOK and exact payloads, see API Service.
| Capability | Quality | Time | # of references | Estimated price |
|---|---|---|---|---|
generate_image | Great | ~10-30s | 16 imgs | $0.07 / $0.10 / $0.15 for 1K / 2K / 4K |
generate_image_pro | Best | ~3-6 min | 32 imgs | $0.26 / $0.45 / $0.77 for 1K / 2K / 4K |
generate_video | Best | ~3-6 min | 9 imgs / 3 vids / 3 auds | $0.08/s / $0.20/s / $0.44/s for 480p / 720p / 1080p |
upscaler | Best | 1-min example: ~4-8 min landed | 1 video | 1-min example: $1.35 for 1080p->4K, $2.35 for 720p->4K |
generate_voice | Good | ~5-15s | N/A | $0.01 per 500 input characters, rounded up |
Resources
- 💬 [Discord][discord-url] — questions, ideas, support, and show & tell
- ⚙️ API Service — developer platform notes, BYOK counterparts, and exact media JSON payloads
- 🎬 Skills reference — how filmmaking skills route agent requests
- 🚀 Setup and agents — choose Docker or host mode, configure Claude/Codex, and understand ports, auth, and permissions
- 🏗 Architecture — viewer, CLI, canvas, and project-file layout
- ❓ FAQ — common setup and generation issues
- 🐛 Issues — bug reports only
- 🤝 Contributing — contribution guide, proprietary-skills carve-out, and CLA flow
License
PAI-Pro is released under the PAI PRO Sustainable Use License, which permits personal use, non-commercial research, and internal business use. Commercial use of PAI-Pro Skills or enterprise-designated source code/Skills requires an explicit agreement; enterprise licenses are available.