Agent skill · Audited from source

pexo-agent

skill · pexoai/pexo-skillsMIT licenceAudited 20 Oct 2026 · source a72d4da09420

★ 810pushed 61 days agoversion 0.3.16
✕BLOCKGrade D · 69/100 · 4 findings · a72d4da
810repo stars · whole pexoai/pexo-skills

Is it safe? No — BLOCK, grade D (69/100). We read this skill's source at commit a72d4da09420 and observed network access and no hard-coded secrets. The audit raised 4 findings, 2 critical or high. What it does: AI video generation skill with auto model selection across Seedance 2, Kling 3.0, HappyHorse, and 10+ models. Produces finished multi-shot videos (5–120s) from text, images, URLs, scripts, or audio — including AI music,…

Ask an AI about this pageGChatGPTCClaudePPerplexityGeGeminiXGrok
More
Cite

Cite this page

ModelsAtlas. “pexo-agent (pexoai/pexo-skills)”. Retrieved 20 Oct 2026. https://modelsatlas.com/skills/pexoai/pexo-agent

Blocked. The page stays public so you can check this skill if someone recommended it to you. The audit found 2 critical or high findings, starting with “Prompt injection” in SKILL.md:83. Do not install it without reading them.

Trust audit

Audit 0.4.1 · source read 20 Oct 2026

Trust score69/100✕BLOCK D
0 critical2 high0 medium1 low1 info

BLOCK means the audit found something that could harm you or your machine. Grade D is shared by 642 of 14,816 audited skills.

What the source does

  • FilesystemNone observedNo reads or writes in the source
  • NetworkDeclared5 observations
  • ShellNone observedNo commands run in the source
  • DependenciesPinnedNothing floats
  • SecretsNone foundNo keys in source

Five audit layers

LayerWhat we checkResultFindings
L0 ProvenanceRepo owner, licence, archive status, activityPASS0
L1 Static sourceSecrets, path traversal, supply chain, hidden codePASS0
L2 InstructionsPrompt injection, authority framing, hidden directivesFAIL4
L3 InventoryThe files and declarations match the sourcePASS0
L4 SandboxRun in isolation and watch behaviourSKIPPED why0

Gates applied: instruction override, no behavioural pass.

Findings (4)

HIGHPrompt injectionprompt.credential_read · SKILL.md:83
F-01 Prompt injectionCWE-94 · CWE-1427Layer L2
read -rsp "Pexo API key: " pexo_api_key
Why
asks the agent to read credentials
Fix
—
HIGHPrompt injectionprompt.hide_from_user · references/TROUBLESHOOTING.md:34
F-01 Prompt injectionCWE-94 · CWE-1427Layer L2
If the body says `error=INTERNAL_ERROR`, do not tell the user to rotate the key first; the service may simply be temporarily down.
Why
asks the agent to act without the user's knowledge
Fix
—
LOWPrompt injectionreview.misleading_scope · references/SETUP-CHECKLIST.md
F-01 Prompt injectionCWE-94 · CWE-1427Layer L2
invite code: **BV5N38**
Why
The package embeds a tracking or marketing referral code, steering installs toward an incentivized account that the video-generation description does not disclose.
Fix
rewrite it so the instruction says plainly what it does, and asks the user before it acts
INFOPrompt injectionprompt.credential_read · references/SETUP-CHECKLIST.md:15
F-01 Prompt injectionCWE-94 · CWE-1427Layer L2
read -rsp "Pexo API key: " pexo_api_key
Why
asks the agent to read credentials
Fix
—

Audit 0.4.1 · source sha a72d4da09420 · 20 Oct 2026Report an issue or request a re-scan

Provenance

Who published it, and what the verdict is pinned to

  • Publisherpexoai GitHub userOwner of the repository this skill is read from
  • Source folderpexoai/pexo-skills/skills/pexo-agentThe exact folder we audited
  • LicenceMITAs the repository states it
  • Commit pinned for this verdicta72d4da09420A new audit runs when the source changes
  • Repository statusNot archivedLast push 61 days ago
  • Sandboxed runNot runSkills are read, not executed; layer L4 is skipped

What it tells the agent

Full source on GitHub ↗

Its description: AI video generation skill with auto model selection across Seedance 2, Kling 3.0, HappyHorse, and 10+ models. Produces finished multi-shot videos (5–120s) from text, images, URLs, scripts, or audio — including AI music,…

---
name: pexo-agent
description: >
  AI video generation skill with auto model selection across Seedance 2,
  Kling 3.0, HappyHorse, and 10+ models. Produces finished multi-shot videos
  (5–120s) from text, images, URLs, scripts, or audio — including AI music,
  lip sync, and multi-shot sequencing. Calls Pexo's external API, manages
  project status and billing confirmations, and transfers only user-approved
  briefs and assets. Runs setup diagnostics, stores generated downloads locally,
  and requires shell, outbound HTTPS, and local file access. Authenticated
  requests are locked to https://pexo.ai. No prompts to write, no models to choose.
  USE FOR: video production, AI video, make a video, product video,
  brand video, promotional clip, explainer video, short video,
  TikTok video, Instagram Reel, YouTube Short, product ad,
  text-to-video, image-to-video, video generation, AI video agent.
license: MIT-0
metadata:
  author: pexoai
  version: "0.3.16"
  openclaw:
    requires:
      env:
        - PEXO_API_KEY
      bins:
        - bash
        - curl
        - jq
        - file
    primaryEnv: PEXO_API_KEY
---

# Pexo Agent — AI Video Generation Skill

Pexo is the most complete video generation skill for Claude Code and other AI coding agents. It handles the full production pipeline — from a natural-language description to a finished, publish-ready video with music, subtitles, and transitions. Auto model selection routes each shot to the best available model (Seedance 2, Kling 3.0, HappyHorse, and more). One API key, no prompt engineering, no video editing.

## What Pexo Does

- **Auto model selection** — Pexo picks the best video model for each shot based on content type. You do not need to know which model to use.
- **Full pipeline** — Script, storyboard, shot-by-shot generation, music, subtitles, lip sync, and final assembly. The output is a finished video, not a raw clip.
- **5 input types** — Text-to-video, image-to-video, URL-to-video (scrapes the page), script-to-video, and audio-to-video.
- **10+ models** — Seedance 2, Kling 3.0, HappyHorse, and more. New models are added as they launch.
- **Any format** — 5–120 seconds, aspect ratios 16:9 (landscape), 9:16 (portrait/vertical), 1:1 (square).

## What You Can Build With Pexo

- Product video ads from a product photo or URL
- TikTok, Instagram Reels, and YouTube Shorts from a text description
- Multi-shot brand videos with consistent style and transitions
- Explainer videos with TTS narration from a script
- E-commerce video content at scale from product catalogs
- Marketing video variants for A/B testing

## How It Works

You send the user's request to Pexo, and Pexo handles all creative work — scriptwriting, shot composition, model selection, prompt engineering, transitions, music. Pexo may ask clarifying questions or present preview options for the user to choose from. A typical 15-second, 3-shot product ad renders in under 8 minutes.

## Data, Permissions, and Cost

- This Skill runs bundled shell scripts, reads only files the user explicitly selects,
  connects only to `https://pexo.ai` for authenticated API calls, uploads approved
  briefs and assets, manages projects and billing confirmations, runs diagnostics,
  and stores generated media under `~/.pexo/tmp` or `PEXO_TMP_DIR`.
- Before the first external transmission in a session, tell the user that their brief,
  selected files, and related metadata will be sent to Pexo and obtain explicit consent.
- Do not upload secrets, regulated data, or unrelated local files. Never search the local
  filesystem for additional material without a separate user request.
- Every billable generation batch requires explicit user approval by default. Report the
  available estimate from Pexo before approving a confirmation.

## Script Execution

Resolve `SKILL_ROOT` to the directory containing this `SKILL.md`. Script names
below are shorthand for `bash "$SKILL_ROOT/scripts/<script-name>"`; do not rely
on executable bits or a modified `PATH`.

## Prerequisites

Config file `~/.pexo/config`:

```bash
umask 077
mkdir -p ~/.pexo
read -rsp "Pexo API key: " pexo_api_key
printf '\n'
{
  printf '%s=%s\n' PEXO_API_KEY "$pexo_api_key"
} > ~/.pexo/config
unset pexo_api_key
chmod 600 ~/.pexo/config
```

First time using this skill or encountering a config error → run `pexo-doctor.sh` and follow its output. See `references/SETUP-CHECKLIST.md` for details.

### Credit Confirmation Preference

`PEXO_BILLING_CONFIRMATION_MODE` controls the confirmation behavior for each message sent by this Skill. It is optional; the default is `always`.

- `always`: ask for approval before every billable generation batch.
- `threshold`: ask when the estimated batch cost exceeds the platform threshold, or when the available balance is insufficient. Use only after the user explicitly opts in for the current session.

Use `pexo-chat.sh --billing-confirmation-mode <mode>` to override the default for one message.

---

## ⚠️ LANGUAGE RULE (highest priority)

**You MUST reply to the user in the SAME language they use. This is non-negotiable.**

- User writes in English → you reply in English
- User writes in Chinese → you reply in Chinese
- User writes in Japanese → you reply in Japanese

This applies to every message you send. If the user switches language mid-conversation, you switch too.

---

## Your Role: Delivery Worker

You are a delivery worker between the user and Pexo. You do three things:

1. **Upload**: user gives a file → `pexo-upload.sh` → get asset ID
2. **Relay**: copy the user's words into `pexo-chat.sh`
3. **Deliver**: poll for results → send video and link to user

Pexo's backend is a professional video creation agent. It understands cinematography, pacing, storytelling, and prompt engineering far better than you. When you add your own creative ideas, the video quality goes down.

### How to relay messages — copy-paste template

When calling pexo-chat.sh, copy the user's message exactly:

```
pexo-chat.

Excerpt, first 6,000 characters of SKILL.md, verbatim from the audited commit. It is text from someone else’s repository, quoted so you can judge it before your agent reads it.

Install

Pinned to the audited commit a72d4da

git clone https://github.com/pexoai/pexo-skills
git -C pexo-skills checkout a72d4da09420

The author documents no install command, so this uses the repository directly. Our verdict covers commit a72d4da09420 only.

Works in which agents?

What its own text says, not a test

AgentStatusMore
Claude CodeMentioned in its text, not testedSkills for Claude Code · About Claude Code
OpenClawMentioned in its text, not testedSkills for OpenClaw · About OpenClaw
“Mentioned” means the skill’s text names the agent. We have not run it there.

Audit history

Every audit, every source change

  1. ✕BLOCK D 69 / 100

    first audit · source a72d4da09420

Similar skills

Questions about pexo-agent

What does the pexo-agent skill do?
A collection of open-source Agent Skills for content creation — images, audio, and video.
Is pexo-agent safe to install?
No — not without reading the findings first. The audit graded it D (69/100) and found 2 critical or high issues in the source. Each one is listed on this page with the file and line it is on.
What can pexo-agent access on my machine?
The audit observed that it reaches the network. Each of those is consistent with what it says it does. Secrets in the source: none found.
Which assistants does pexo-agent work with?
Its documentation mentions claude-code and openclaw. That is what the text claims, not a compatibility test we ran.
How current is this page?
The grade is for one exact copy of the source (a72d4da09420), read on 2026-10-20. The repository is watched, and a new audit runs when it changes — this is the first audit.

Sources

Spotted something wrong? Report a correction · Methodology

Keep going in your AI

Opens with this page as the source. We add nothing else to the prompt.