Resemble Detect skill
Resemble AI skill that checks audio, video, images and text for synthesis; verdicts come only from completed jobs.
- 103 ★ on the repository
- Last commit
- Apache-2.0
Extended access
Why this level
- Files and links to material are sent to the external Resemble AI service.
- The API key lives in an environment variable and is billed by usage.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/resemble-detectDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Files and links to material are sent to the external Resemble AI service.
- The API key lives in an environment variable and is billed by usage.
This is third-party code. Review the repository files before installing.
What it does
The skill teaches an agent to call the Resemble Detect API with `curl`. The agent submits audio, an image or a video by URL or uploads a file up to 150 MB, then reads the label, score and status of the job. It also covers source tracing for synthesized voices, speaker, emotion and transcript analysis, AI-authorship checks for text, and counting website visitors that turn out to be AI agents. The skill's core rule is that the agent never calls media real or fake until a detection job has completed. While processing runs it waits, on failure it says so, and it never replaces the result with its own guess.
Who it is for. Moderators, journalists and security analysts who verify the authenticity of recordings and attachments through Claude Code, Cursor or Gemini CLI.
Good fit when
- A newsroom received an audio recording and wants to know whether the voice is synthesized before publishing.
- An insurer received a damage photo and needs a verdict plus a map of suspicious areas.
- A security team checks which service may have voiced a suspicious call.
- A site owner wants to know what share of visits comes from AI agents and what share from people.
Not a fit when
- You need to generate speech or clone a voice; the skill only checks.
- The material cannot be sent to an external service.
- There is no Resemble AI key and the check must run offline.
Example request
Check whether the voice in the recording at this link is synthesized, and show the label, score and the likely voice service.Limitations
A Resemble AI API key is required and is billed under the service plan. Files and links go to Resemble servers; the request has a zero-retention mode, but it must be enabled explicitly. Text checks accept 25 words or more, and the first request after idle takes several minutes. The service verdict is probabilistic: errors are possible both ways, and a person decides on important material. The skill does no transcription, speech generation or voice cloning.
How to disable. Delete the skill folder from `.claude/skills/` or another agent directory and revoke the key in the Resemble AI dashboard.
FAQ
How do I install the skill in Claude Code?
Run `npx skills add resemble-ai/detect-skill` or copy `SKILL.md` into `.claude/skills/`. For Cursor the file goes into `.cursor/skills/`.
Is an MCP server required?
No. The skill calls the REST API through `curl`. The Resemble MCP server can be added for live documentation of the endpoints.
Which files can be checked?
Audio, images and video by a public HTTPS URL or an upload up to 150 MB. Larger and private files use a secure upload token.
Can it tell whether text was written by AI?
Yes, for text of 25 words or more the service returns `ai` or `human` and a confidence value. The agent relays it without judging style itself.
How does Agent Detection differ from Detect Agents?
Agent Detection tells whether a site visitor is a person or an AI agent. Detect Agents runs a multi-step investigation over material such as a document or a claim.
Discussion
No questions or reviews yet.
Related
SkillsA security scanner for agent skills: finds prompt injection, data exfiltration and malicious code before you install
A code security audit skill by Cloudflare: the agent runs recon, coverage-led hunting and independent verification of findings, then produces a structured repor
Cybersecurity Skills
Cybersecurity Skills for AI Agents
42 security skills for Claude Code, Cursor and Codex: code and cloud audits, authorization-gated pentesting and incident response
Security and 152-FZ audit skill with rouble risk pricing
axioma-security
A skill for a deep website security and 152-FZ compliance audit: every finding is mapped to a law article and priced in roubles
VibeSec
VibeSec-Skill
A secure coding skill: the agent looks at code from a bug hunter's perspective and prevents common web vulnerabilities
Twenty Didit skills for identity checks: the agent runs KYC, sanctions screening and biometric checks through the API
README in short
The skill is a single `SKILL.md` file: a decision tree for the request, ready `curl` examples, rules for reading scores, workflow templates, a list of typical agent mistakes and a table of API response codes.
README badge
Original SKILL.md
---
name: resemble-detect
description: Deepfake detection and media safety — detect AI-generated audio, images, video, and text, trace synthesis sources, analyze media intelligence, and tell people from AI agents on a website (Agent Detection) using direct Resemble AI API calls
---
# Resemble Detect — Deepfake Detection & Media Safety
Analyze audio, image, video, and text for synthetic manipulation, AI-generated content, and media intelligence, and find out whether a website's visitors are people or AI agents, using **direct Resemble AI API calls**.
## Core Principle — THE IRON LAW
**"NEVER DECLARE MEDIA AS REAL OR FAKE WITHOUT A COMPLETED DETECTION RESULT."**
Do not guess, infer, or speculate about media authenticity. Every authenticity claim must be backed by a completed Resemble Detect job with a returned `label`, `score`, and `status: "completed"`. If the detection is still `processing`, wait. If it `failed`, say so — do not substitute your own judgment.
The same law applies to text. Never call writing AI-generated or human-written from style, tone, or "it reads like ChatGPT." A text verdict requires a completed `POST /text_detect` job with `prediction`, `confidence`, and `status: "completed"`.
The same law applies to website visitors. Never say a site's traffic is agents or people from user agents, request volume, or intuition. A visitor verdict comes from Agent Detection (`settled` on a visit, or the totals from `GET /agent_detection/analytics`).
## When to Use
Use this skill whenever the user's request involves any of these:
- Checking if audio, video, or image is AI-generated or manipulated
- Checking if text — an essay, email, post, review, article, cover letter, or comment — was written by an AI model ("slop detection")
- Detecting deepfakes in any media format
- Verifying media authenticity or provenance
- Identifying which AI platform synthesized audio (source tracing)
- Analyzing media for speaker info, emotion, transcription, or misinformation
- Asking natural-language questions about detection results
- Running a full investigation workflow — insurance claim, breaking news, ID check, submitted evidence — where detection is one input among several
- Finding out whether a website's visitors are people or AI agents, adding Agent Detection to a site, or reading how much of a site's traffic comes from agents
- Any mention of: "deepfake", "fake detection", "synthetic media", "media forensics", "authenticity check", "source tracing", "is this real", "AI-written", "written by ChatGPT", "AI text", "slop", "is this human-written", "agent detection", "bot traffic", "AI agents visiting my site", "Porter", "is this visitor a bot"
**Do NOT use** for text-to-speech generation, voice cloning, or speech-to-text transcription — those are separate Resemble capabilities. (Detecting whether *text* is AI-written **is** in scope — see Phase 5. Telling people from AI agents on a website **is** in scope — see Phase 6.)
**Agent Detection is not Detect Agents.** Detect Agents (Phase 4) are Resemble's investigators that analyze media. Agent Detection (Phase 6) tells a website whether each *visitor* is a person or an AI agent.
## Required Setup