Back to skill

Security audit

Screencast Studio

Security checks for vulnerabilities and agentic risk

Overview

The skill is a coherent browser-demo recorder, but it needs review because it persists login session files and ships an automatic cleanup script that can delete unrelated workspace files.

Install only in a fresh, dedicated demo directory, not an existing project. Use a test account and staging data where possible, avoid remote production login while TLS verification is disabled, keep storageState.json and post-login artifacts out of source control, delete or revoke the session after recording, and review clean.js before running npm run ship.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
Findings (5)

T09 · Insecure Skill Coding Practices

Error
Location
templates/clean.js:20
Finding

Destructive Allowlist-Based Cleanup Can Delete User Projects and Symlink Targets

Content
View full analysis
rmRecurse(path.join(p, f))); fs.rmdirSync(p); } else { fs.unlinkSync(p); } } let removed = 0; for (const name of fs.readdirSync(ROOT)) { const p = path.join(ROOT, name); const isDir = fs.statSync(p).isDirectory(); if (isDir && KEEP_DIRS.has(name)) continue; if (!isDir && KEEP_FILES.has(name)) continue; console.log('rm', isDir ? `${name}/` : name); rmRecurse(p); removed++; } ``` The cleanup script is automatically executed after a successful ship: ```json "ship": "node record.js && node postprocess.js && node deploy.js && node review.js && node clean.js" ``` ### Technical Analysis The cleanup implementation treats every file or directory not included in a small allowlist as disposable. The Skill documentation instructs users to copy the templates into a user-selected working directory, which may be an existing application repository rather than a dedicated generated directory. Consequently, normal application files such as source directories, configuration files, documentation, tests, and version-control metadata can be recursively deleted when `npm run ship` reaches its cleanup phase. The use of `fs.statSync()` introduces an additional boundary issue because it follows symbolic links. If an unallowlisted entry is a symbolic link to a directory outside the workspace, the recursive function may traverse and delete content from the linked target rather than deleting only the link itself. ### Attack Path 1. The templates are copied into an existing application or repository directory. 2. The directory contains source files or folders not present i ...[truncated 999 chars]
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
templates/login.js:23
Finding

TLS Certificate Verification Is Disabled During Authentication and Recording

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
templates/login.js:45
Finding

Reusable Authentication State and Post-Login Data Are Retained as Unprotected Files

Content
View full analysis
{ const ts = (el) => (el.innerText || '').trim().replace(/\s+/g, ' ').slice(0, 100); const navItems = [...document.querySelectorAll('nav a, [role="navigation"] a, aside a, [class*="sidebar"] a, [class*="menu"] a')] .slice(0, 30) .map((el) => ({ text: ts(el), href: el.getAttribute('href') })); const headings = [...document.querySelectorAll('h1,h2,h3')] .slice(0, 20) .map((el) => ({ tag: el.tagName, text: ts(el) })); const buttons = [...document.querySelectorAll('button')] .slice(0, 30) .map((el) => ({ text: ts(el), title: el.title })) .filter((b) => b.text); return { url: location.href, navItems, headings, buttons }; }); require('fs').writeFileSync(path.join(__dirname, 'post-login-summary.json'), JSON.stringify(summary, null, 2)); await context.storageState({ path: path.join(__dirname, 'storageState.json') }); ``` The cleanup script deliberately preserves these artifacts: ```js // session + assets 'storageState.json', 'cursor.png', 'ripple.png', 'post-login.png', 'post-login-summary.json', ``` ### Technical Analysis Playwright storage-state files can contain cookies and origin storage sufficient to restore an authenticated session. The script writes this data to `storageState.json` without explicitly applying restrictive file permissions. The script also captures a screenshot and a text summary immediately after authentication. Those artifacts can contain private navigation labels, project names, int ...[truncated 1629 chars]
Remediation
View remediation

T08 · Insecure Dependencies

Warning
Location
templates/package.json:22
Finding

Unpinned Dependency Graph Is Installed Without a Supplied Lockfile

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Note
Location
templates/deploy.js:7
Finding

Configurable Deployment Prefix Allows Destination Path Traversal

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
Findings (38)

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill goes beyond simple screencast generation by instructing manual login, storage state capture, DOM text extraction, and post-login screenshots. Those behaviors handle authenticated session material and potentially sensitive application data, which materially changes the security profile from a harmless recording tool to one that can collect secrets and internal information.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The skill goes beyond simple screencast generation by instructing manual login, storage state capture, DOM text extraction, and post-login screenshots. Those behaviors handle authenticated session material and potentially sensitive application data, which materially changes the security profile from a harmless recording tool to one that can collect secrets and internal information.

Content

No source excerpt is available for this finding.

Vague Triggers

High
Category
Not specified by scanner
Confidence
95% confidence
Finding

The trigger list includes a broad catch-all 'Anything where the user said' followed by phrases such as '给同事看看' and 'walkthrough video,' mixing specific recording intents with common expressions that could arise in everyday collaboration. This increases the chance of activating the skill for general requests to show work to a teammate rather than explicit requests to create a screencast.

Content

No source excerpt is available for this finding.

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 66)May include surrounding context.

md
templates/record.js → <working-dir>/record.js

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 95)May include surrounding context.

md
templates/record.js → <working-dir>/record.js

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 104)May include surrounding context.

md
templates/record.js → <working-dir>/record.js

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 123)May include surrounding context.

md
templates/record.js → <working-dir>/record.js

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 141)May include surrounding context.

md
templates/record.js → <working-dir>/record.js

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 178)May include surrounding context.

md
templates/record.js → <working-dir>/record.js

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 67)May include surrounding context.

md
templates/postprocess.js → <working-dir>/postprocess.js

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 68)May include surrounding context.

md
templates/review.js → <working-dir>/review.js

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 72)May include surrounding context.

md
templates/deploy.js → <working-dir>/deploy.js

Anti-Refusal Statement

High
Category
Anti-Refusal
Confidence
85% confidence
Finding

Skill instructs the agent to omit warnings, disclaimers, or ethical commentary. Stripping safety caveats hides risk from the user and is a common jailbreak preamble.

Content

Scanner excerpt · SKILL.md (reported line 137)May include surrounding context.

md
If the user says "no masks needed" (e.g. demo is on a generic public site with no PII), leave the array empty — the recording proceeds without any blur. Don't lecture.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

Saving Playwright storage state to disk can preserve cookies, local storage, and other authentication material that may allow session reuse if the file is accessed by another local user, malware, or accidentally committed to source control. This effectively creates a bearer-like credential artifact without a strong warning or protective controls. In this skill, the helper intentionally records a logged-in session for later automation, so the context makes the persisted credential especially sensitive and directly exploitable.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The example invocation is presented as a direct command to the assistant in Chinese ('用 screencast-studio 给我录个 demo') with no indication that other languages are supported. This can constitute a language/locale policy issue because it implicitly requires a specific language without user opt-in or alternatives.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
91% confidence
Finding

The skill instructs the agent to use environment variables and execute setup/login/ship commands, but it does not declare any explicit tool or permission scope. That creates an authorization gap where an agent may overreach into shell/env access without the user or platform having a clear least-privilege boundary.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The manifest description says to activate when the user wants to 'test or share something they just vibe-coded' or asks for a 'polished walkthrough,' which is broad and may overlap with ordinary requests for feedback, testing, or explanation rather than screencast creation. Although examples are provided elsewhere, this top-level trigger language is still ambiguous enough to cause unintended invocation.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

Using npx playwright install chromium without pinning a version allows retrieval of whatever package version resolves at execution time. This weakens supply-chain integrity and can lead to unexpected or malicious code being executed if the dependency changes or is compromised.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
82% confidence
Finding

The helper examples and authoring guidance use Chinese narration text such as '操作描述' and '进入项目模块' as the model pattern, but the document does not explicitly tell the agent to match the user's preferred language. That can lead to a locale policy violation if the skill defaults to Chinese output for users who did not opt in.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

Referencing npx playwright again in prerequisites perpetuates the same unpinned remote package execution risk. Repeated instructions increase the chance operators will run an unverified version in different environments.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

This example explicitly instructs the automation to create a timestamped folder and upload files into the target application, which modifies application data. In the context of a screencast/testing skill, that behavior may be expected, but without an explicit warning, consent step, or strong guidance to use only test/staging data, users may run it against production systems and unintentionally pollute real datasets or trigger retention/compliance issues.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
87% confidence
Finding

The documentation explicitly recommends page.on('dialog', (d) => d.accept()), which will automatically confirm every browser dialog, including destructive confirmations such as delete, sign-out, reset, or permission prompts. In a Playwright-driven recording skill that interacts with arbitrary web UIs, this increases the chance of unintended state changes and can silently bypass a human review step that the dialog was meant to enforce.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
70% confidence
Finding

npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
70% confidence
Finding

npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
70% confidence
Finding

npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.

Content

No source excerpt is available for this finding.

Static analysis

Detected: suspicious.dangerous_exec

Shell command execution detected (child_process).

Critical
Code
suspicious.dangerous_exec
Location
templates/postprocess.js:136

Shell command execution detected (child_process).

Critical
Code
suspicious.dangerous_exec
Location
templates/review.js:46