Skills/Glama/backdraw

backdraw@2.4.0

CAUTION· static only

Publisher claims: "Lets Claude and other MCP clients generate and edit images with your own signed-in ChatGPT account in a background Chrome tab. No API key, no open ports; right-click reference collections and a gallery side panel. Windows + Chrome." Observed in sandbox: static analysis only — behavior not yet observed.

Registry GlamaPublisher devtalh · publishing for 0 daysInstalls 0Last detonated 8 h ago
What would fix thisWatchDispute verdict

Coverage

Static only; nothing was run yet.

The label rests on the code, the instruction files and the publisher. The honeypot run replaces this verdict when it completes.

Before you give it your token

It will ask for
no credential variables read in the code
Where it talks to
no vendor host named in the codeAlso chatgpt.com — not the publisher's domain; check what for before connecting.
What the honeypot saw
nothing has run yet
What we could not test
everything: the server has not run yet

Run it before you trust it. Nothing has run yet: the code was read, the server was not. `vet-skills try` runs this exact version in the honeypot in about a minute.

Never paste a credential into a chat, a ticket or a report, including ours. How the guard holds it.

Is [email protected] safe to install?

KillerSkills analysed it statically (a sandbox run is pending) on 2026-11-14 and scored it 60/100 (Caution). It is not known to be malicious, but it has behaviour or code a security policy should decide on before install. Notable findings: Hidden zero-width characters detected. Summary: static analysis only — behavior not yet observed.

Behavior trace

Queued for honeypot run · static analysis only
t+CategoryObserved action

This version has not been detonated yet. The score above counts behavior as unknown (50/100), so it cannot read Verified until the honeypot run completes and replaces this verdict.

Static findings

MEDIUM
npx invocation of non-well-known packagedoctor.js:132 · npx invocation of non-well-known package — may execute untrusted code// Only an entry that can't work is rewritten; an npx entry or another working copy never is.Fix: Pin versions and sources; vendor anything fetched at install time.
AST06
MEDIUM
npx invocation of non-well-known package.claude-plugin/marketplace.json:12 · npx invocation of non-well-known package — may execute untrusted code"description": "The Backdraw MCP server (runs npx backdraw) and /backdraw:setup for the one-time browser setup.",Fix: Pin versions and sources; vendor anything fetched at install time.
AST06
MEDIUM
npx invocation of non-well-known packagebin/backdraw.js:4 · npx invocation of non-well-known package — may execute untrusted code// npx backdraw@latest setupFix: Pin versions and sources; vendor anything fetched at install time.
AST06
MEDIUM
npx invocation of non-well-known packagelib/doctor.js:9 · npx invocation of non-well-known package — may execute untrusted codeconst SETUP = 'npx backdraw@latest setup';Fix: Pin versions and sources; vendor anything fetched at install time.
AST06
HIGH
Hidden zero-width characters detectedlib/mcp-clients.js:91 · Hidden zero-width characters detected — may conceal contentconst text = fs.readFileSync(file, 'utf8').replace(/^/, '');Fix: Keep SKILL.md to visible, descriptive text. Remove comments and encoded text that address the agent.
AST07
MEDIUM
npx invocation of non-well-known packagelib/mcp-clients.js:169 · npx invocation of non-well-known package — may execute untrusted code// The package an npx entry runs: `npx backdraw@latest mcp`, also through `cmd /c` (how WindowsFix: Pin versions and sources; vendor anything fetched at install time.
AST06
MEDIUM
File type combined with upload actiontest/host.test.js:36 · File type combined with upload action — agent may exfiltrate local data filesfs.writeFileSync(path.join(DIR, 'data', 'config.json'), JSON.stringify({ outputDir: IMAGES, collectionsDir: REFS, project: PROJECT }));Fix: Remove the call, or declare the host in frontmatter and explain in the README what is sent and why.
AST07
LOW
Fake instruction boundary tagREADME.md:674 · Fake instruction boundary tag · informational: not an instruction filenode cli.js generate "<prompt>" [--ref IMAGE]... [--url CHAT_LINK] [--aspect W:H] [--transparent]Fix: Remove the text that addresses the agent rather than the reader; describe behavior, do not instruct.
AST01
LOW
Prompt injectionREADME.tr.md:24 · Prompt injection — known jailbreak technique reference · informational: not an instruction fileBackdraw, bilgisayarında çalışan bir MCP sunucusu ve bir Chrome eklentisinden oluşur. Claude'danFix: Keep SKILL.md to visible, descriptive text. Remove comments and encoded text that address the agent.
AST01
LOW
Prompt injectionREADME.tr.md:15 · Prompt injection — known jailbreak technique reference · informational: not an instruction file…ki görseller sağ tıkla referans olarak kaydediliyor, sonra Claude Code Backdraw'dan o tarzda bir görsel istiyor; ChatGPT görseli arka plan sekmesinde çiziyor ve do…Fix: Keep SKILL.md to visible, descriptive text. Remove comments and encoded text that address the agent.
AST01
MEDIUM
npx invocation of non-well-known packagetest/doctor.test.js:70 · npx invocation of non-well-known package — may execute untrusted codeassert.deepEqual([by('Cursor')[0].status, by('Cursor')[0].fix], ['warn', 'Run: npx backdraw@latest setup --clients=cursor']);Fix: Pin versions and sources; vendor anything fetched at install time.
AST06
LOW
Prompt injectionREADME.md:194 · Prompt injection — known jailbreak technique reference · informational: not an instruction file3. **Load the extension** once: open `chrome://extensions`, turn on **Developer mode**, clickFix: Keep SKILL.md to visible, descriptive text. Remove comments and encoded text that address the agent.
AST01
LOW
Prompt injectionREADME.tr.md:231 · Prompt injection — known jailbreak technique reference · informational: not an instruction file### Claude Code'danFix: Keep SKILL.md to visible, descriptive text. Remove comments and encoded text that address the agent.
AST01
LOW
Prompt injectionREADME.md:146 · Prompt injection — known jailbreak technique reference · informational: not an instruction file1. turn on **Developer mode** (top right),Fix: Keep SKILL.md to visible, descriptive text. Remove comments and encoded text that address the agent.
AST01
MEDIUM
File type combined with upload actiontest/install.test.js:82 · File type combined with upload action — agent may exfiltrate local data filesassert.deepEqual(readJson(path.join(newDir, 'data', 'config.json')), { outputDir: 'C:\\Pictures\\Old' });Fix: Remove the call, or declare the host in frontmatter and explain in the README what is sent and why.
AST07
MEDIUM
File type combined with upload actiontest/install.test.js:282 · File type combined with upload action — agent may exfiltrate local data fileswriteJson(path.join(legacy, 'data', 'config.json'), { outputDir: path.join(t.home, 'OldPics') });Fix: Remove the call, or declare the host in frontmatter and explain in the README what is sent and why.
AST07
MEDIUM
npx invocation of non-well-known packagesetup.js:2 · npx invocation of non-well-known package — may execute untrusted code// `backdraw setup` (npx backdraw@latest setup): installs Backdraw into %USERPROFILE%\.backdraw (npx'sFix: Pin versions and sources; vendor anything fetched at install time.
AST06
LOW
Authorization escalation languageREADME.md:194 · Authorization escalation language · informational: not an instruction file3. **Load the extension** once: open `chrome://extensions`, turn on **Developer mode**, clickFix: Remove the text that addresses the agent rather than the reader; describe behavior, do not instruct.
AST01
MEDIUM
npx invocation of non-well-known packagelib/doctor.js:10 · npx invocation of non-well-known package — may execute untrusted codeconst FIX = 'npx backdraw@latest doctor --fix';Fix: Pin versions and sources; vendor anything fetched at install time.
AST06
MEDIUM
File type combined with upload actiontest/install.test.js:284 · File type combined with upload action — agent may exfiltrate local data filesassert.equal(readJson(path.join(t.dest, 'data', 'config.json')).outputDir, path.join(t.home, 'OldPics'));Fix: Remove the call, or declare the host in frontmatter and explain in the README what is sent and why.
AST07
MEDIUM
npx invocation of non-well-known packagetest/doctor.test.js:75 · npx invocation of non-well-known package — may execute untrusted codeassert.match(text, /\n7 problems, 2 warnings\. "npx backdraw@latest doctor --fix" can repair 3 of them\.$/);Fix: Pin versions and sources; vendor anything fetched at install time.
AST06
MEDIUM
File type combined with upload actiontest/host.test.js:857 · File type combined with upload action — agent may exfiltrate local data filesassert.equal(fs.readFileSync(path.join(collision, 'keep.txt'), 'utf8'), 'leave existing outputs alone');Fix: Remove the call, or declare the host in frontmatter and explain in the README what is sent and why.
AST07
LOW
Authorization escalation languageREADME.md:146 · Authorization escalation language · informational: not an instruction file1. turn on **Developer mode** (top right),Fix: Remove the text that addresses the agent rather than the reader; describe behavior, do not instruct.
AST01
MEDIUM
npx invocation of non-well-known packagelib/doctor.js:92 · npx invocation of non-well-known package — may execute untrusted code// fixable; an npx entry, another working copy or an entry it doesn't know is reported and left alone.Fix: Pin versions and sources; vendor anything fetched at install time.
AST06
MEDIUM
npx invocation of non-well-known packagetest/doctor.test.js:69 · npx invocation of non-well-known package — may execute untrusted codeassert.equal(by('Claude Code')[0].fix, 'npx backdraw@latest doctor --fix');Fix: Pin versions and sources; vendor anything fetched at install time.
AST06
LOW
Prompt injectionREADME.tr.md:367 · Prompt injection — known jailbreak technique reference · informational: not an instruction file3. Claude'dan koleksiyonu adıyla kullanmasını iste.Fix: Keep SKILL.md to visible, descriptive text. Remove comments and encoded text that address the agent.
AST01
MEDIUM
npx invocation of non-well-known packagesetup.js:8 · npx invocation of non-well-known package — may execute untrusted code// An entry that starts Backdraw through npx stays as it is (npx fetches the new version itself).Fix: Pin versions and sources; vendor anything fetched at install time.
AST06
MEDIUM
File type combined with upload actiontest/host.test.js:852 · File type combined with upload action — agent may exfiltrate local data filesfs.writeFileSync(path.join(collision, 'keep.txt'), 'leave existing outputs alone');Fix: Remove the call, or declare the host in frontmatter and explain in the README what is sent and why.
AST07
MEDIUM
npx invocation of non-well-known packagelib/doctor.js:96 · npx invocation of non-well-known package — may execute untrusted codeelse add('ok', label, e.kind === 'npx' ? `set up via npx (${e.spec})` : 'set up for this install');Fix: Pin versions and sources; vendor anything fetched at install time.
AST06
MEDIUM
File type combined with upload actiontest/install.test.js:77 · File type combined with upload action — agent may exfiltrate local data fileswriteJson(path.join(oldDir, 'data', 'config.json'), { outputDir: 'C:\\Pictures\\Old' });Fix: Remove the call, or declare the host in frontmatter and explain in the README what is sent and why.
AST07
MEDIUM
npx invocation of non-well-known packagesetup.js:94 · npx invocation of non-well-known package — may execute untrusted code// A file whose entry starts npx is kept as it is.Fix: Pin versions and sources; vendor anything fetched at install time.
AST06
MEDIUM
npx invocation of non-well-known packagetest/doctor.test.js:65 · npx invocation of non-well-known package — may execute untrusted codeassert.deepEqual([by('Version')[0].status, by('Version')[0].fix], ['warn', 'Update: npx backdraw@latest setup']);Fix: Pin versions and sources; vendor anything fetched at install time.
AST06
MEDIUM
npx invocation of non-well-known packagetest/doctor.test.js:74 · npx invocation of non-well-known package — may execute untrusted codeassert.match(text, /✗ Google Chrome +the native host is not registered\n +→ npx backdraw@latest doctor --fix\n/);Fix: Pin versions and sources; vendor anything fetched at install time.
AST06
MEDIUM
npx invocation of non-well-known packageserver.json:31 · npx invocation of non-well-known package — may execute untrusted code"requirements": "Windows 10/11, Google Chrome 116+, Node.js 20+, a ChatGPT account signed in in Chrome. Run \"npx backdraw@latest setup\" once to install the browser extension and native host.",Fix: Pin versions and sources; vendor anything fetched at install time.
AST06
MEDIUM
npx invocation of non-well-known packagesetup.js:54 · npx invocation of non-well-known package — may execute untrusted code// The npx package a client's entry runs ("backdraw@latest"), or null when it starts a local file.Fix: Pin versions and sources; vendor anything fetched at install time.
AST06
LOW
Prompt injectionREADME.tr.md:439 · Prompt injection — known jailbreak technique reference · informational: not an instruction fileEklenen önizleme tüm kareleri numaralı bir ızgarada gösterir (16'dan fazlaysa eşit aralıklı 16 kare);Fix: Keep SKILL.md to visible, descriptive text. Remove comments and encoded text that address the agent.
AST01
MEDIUM
npx invocation of non-well-known packageplugin/skills/setup/SKILL.md:19 · npx invocation of non-well-known package — may execute untrusted codework, `npx backdraw@latest doctor` checks every piece and says how to fix it.Fix: Pin versions and sources; vendor anything fetched at install time.
AST06
MEDIUM
npx invocation of non-well-known packageplugin/skills/setup/SKILL.md:11 · npx invocation of non-well-known package — may execute untrusted codenpx backdraw@latest setup --clients=noneFix: Pin versions and sources; vendor anything fetched at install time.
AST06

Cost to the agent

Instructions loaded
≈269 tok
1 file · 80 in bundle

Tokens estimated at four characters each. Skills load their instruction text when triggered; the figure is what that costs.

Known vulnerabilities

Vulnerability advisories (OSV) cover npm and PyPI packages; this registry has no advisory database.

What would make this Verified

20 points to 80 · weights: behavior 45%, static 25%, prompt 15%, publisher 15%
  1. Unclaimed publisher≈ +8 on the totalClaim devtalh: a verified publisher gets +50 on publisher trust. It is free and usually instant with GitHub sign-in.
  2. Hidden zero-width characters detected≈ +4 on the totalKeep SKILL.md to visible, descriptive text. Remove comments and encoded text that address the agent.
  3. npx invocation of non-well-known package≈ +2 on the totalPin versions and sources; vendor anything fetched at install time.
  4. File type combined with upload action≈ +2 on the totalRemove the call, or declare the host in frontmatter and explain in the README what is sent and why.
Publish a new version to Glama, then queue it here. Not your skill? Claim the handle to get verdict changes by email.Claim devtalhPropose a patch to the maintainer