Back to skill

Security audit

website-building

Security checks for vulnerabilities and agentic risk

Overview

This skill transparently builds and deploys a Cargo-hosted static company website, with DNS and deploy risks disclosed and gated on review.

Install only if you are comfortable letting the workflow modify website source, plan Cargo resources, and guide DNS changes. Review the Cargo plan before any deploy, require explicit approval for domain purchases or DNS replacements, and update or audit the pinned web dependencies and any @latest tooling before using the generated site in production.

Vulnerability Patterns
  • 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
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
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
  • Tool MisuseTool Parameter Abuse, Chaining Abuse, Unsafe Defaults
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (16)

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 100)May include surrounding context.

md
6. **Plan, then stop.** `node --import tsx evals/contract.mjs` from this skill's folder, then

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 126)May include surrounding context.

md
6. **Plan, then stop.** `node --import tsx evals/contract.mjs` from this skill's folder, then

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 169)May include surrounding context.

md
6. **Plan, then stop.** `node --import tsx evals/contract.mjs` from this skill's folder, then

Known Vulnerable Dependency: next==16.3.7 — 6 advisory(ies): CVE-2026-94486 (Next.js has information disclosure in development server's Model Context Protoco); CVE-2026-94544 (Next.js: Pending `use cache` fill can leak Draft Mode content into regular respo); CVE-2026-94543 (Next.js has cache poisoning of SSG and ISR pages in self-hosted applications) +3 more

High
Category
Supply Chain
Confidence
97% confidence
Finding

The lockfile pins Next.js 16.3.7, and the reported advisories include cache poisoning, draft-mode leakage, and development-server information disclosure issues. Even though this is a marketing website skill and some advisories may depend on deployment mode or enabled features, shipping a known-vulnerable framework version is a real supply-chain risk because the app could be self-hosted and exposed on a public domain.

Content

No source excerpt is available for this finding.

Known Vulnerable Dependency: source-map-js==1.2.1 — 1 advisory(ies): CVE-2026-93749 (source-map-js allows event-loop denial of service through indexed source-map sec)

High
Category
Supply Chain
Confidence
83% confidence
Finding

source-map-js 1.2.1 is present and is flagged for an event-loop denial-of-service condition when processing malicious indexed source maps. In this website skill, the package appears as a transitive build/tooling dependency, so exploitability is narrower than a runtime web-facing library, but it is still a real dependency risk if attacker-controlled source maps are ever processed in build, analysis, or debugging workflows.

Content

No source excerpt is available for this finding.

Chaining Abuse

High
Category
Tool Misuse
Confidence
75% confidence
Finding

Tool calls are chained to bypass individual safety checks or escalate capabilities beyond what any single tool call would allow.

Content

Scanner excerpt · infra/apps/website/package.json (reported line 9)May include surrounding context.

json
"scripts": {
    "dev": "next dev --hostname 127.0.0.1",
    "check": "tsc --noEmit",
    "build": "next build && rm -rf dist && mv out dist"
  },
  "dependencies": {
    "class-variance-authority": "0.7.1",

Known Vulnerable Dependency: next==16.3.7 — 6 advisory(ies): CVE-2026-94486 (Next.js has information disclosure in development server's Model Context Protoco); CVE-2026-94544 (Next.js: Pending `use cache` fill can leak Draft Mode content into regular respo); CVE-2026-94543 (Next.js has cache poisoning of SSG and ISR pages in self-hosted applications) +3 more

High
Category
Supply Chain
Confidence
97% confidence
Finding

The package pins Next.js 16.3.7, and the provided advisory list includes multiple high-severity issues affecting that version, including information disclosure and cache-related content leakage/poisoning. In the context of a website-building skill intended to publish a public site, shipping a known-vulnerable web framework is a real supply-chain risk because the generated site or associated build/runtime environment may expose sensitive content or serve poisoned responses if deployed in an affected mode.

Content

No source excerpt is available for this finding.

Whitespace Padding

Medium
Category
Prompt Injection
Confidence
70% confidence
Finding

Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.

Content

Scanner excerpt · SKILL.md (reported line 113)May include surrounding context.

md
**Derive before you ask.** An input with a lookup is looked up, not asked.

| Input                                                                              | Kind  | How it is answered                                                                                                                                     | Why it matters                                                                                                                              |
| ---------------------------------------------------------------------------------- | ----- | ------------------------------------------------------------------------------------------------------------------------------------------------------ | ------------------------------------------------------------------------------------------------------------------------------------------- |
| company, audience, offering, proof                                                 | value | **derived**: `context/` and the existing site or brand guide; ask only for what is missing                                                             | The pages say only what the company approved. Placeholder copy and invented proof never ship.                                               |
| pages, call to action, design direction                                            | asked | propose a brief from the context and let the operator correct it                                                                                       | The brief is the scope every later pull request is reviewed against.                                                                        |

Whitespace Padding

Medium
Category
Prompt Injection
Confidence
70% confidence
Finding

Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.

Content

Scanner excerpt · SKILL.md (reported line 113)May include surrounding context.

md
**Derive before you ask.** An input with a lookup is looked up, not asked.

| Input                                                                              | Kind  | How it is answered                                                                                                                                     | Why it matters                                                                                                                              |
| ---------------------------------------------------------------------------------- | ----- | ------------------------------------------------------------------------------------------------------------------------------------------------------ | ------------------------------------------------------------------------------------------------------------------------------------------- |
| company, audience, offering, proof                                                 | value | **derived**: `context/` and the existing site or brand guide; ask only for what is missing                                                             | The pages say only what the company approved. Placeholder copy and invented proof never ship.                                               |
| pages, call to action, design direction                                            | asked | propose a brief from the context and let the operator correct it                                                                                       | The brief is the scope every later pull request is reviewed against.                                                                        |

Whitespace Padding

Medium
Category
Prompt Injection
Confidence
70% confidence
Finding

Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.

Content

Scanner excerpt · SKILL.md (reported line 116)May include surrounding context.

md
| Input                                                                              | Kind  | How it is answered                                                                                                                                     | Why it matters                                                                                                                              |
| ---------------------------------------------------------------------------------- | ----- | ------------------------------------------------------------------------------------------------------------------------------------------------------ | ------------------------------------------------------------------------------------------------------------------------------------------- |
| company, audience, offering, proof                                                 | value | **derived**: `context/` and the existing site or brand guide; ask only for what is missing                                                             | The pages say only what the company approved. Placeholder copy and invented proof never ship.                                               |
| pages, call to action, design direction                                            | asked | propose a brief from the context and let the operator correct it                                                                                       | The brief is the scope every later pull request is reviewed against.                                                                        |
| source                                                                             | asked | a new site, a recreation of an authorized repository or live site, or a redesign                                                                       | A supplied repository is built from; a screenshot is no substitute for it.                                                                  |
| domain and who holds its DNS (`infra/apps/websit
...[truncated 25 chars]

Whitespace Padding

Medium
Category
Prompt Injection
Confidence
70% confidence
Finding

Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.

Content

Scanner excerpt · SKILL.md (reported line 133)May include surrounding context.

md
The code is a worked example. These reshapes are expected, and the agent offers them rather than
waiting to be asked. Every one costs something.

| Variation           | When it is right                                                              | How                                                                                                                                                   | What it costs                                                                                                |
| ------------------- | ----------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------ |
| `external-dns`      | The domain's DNS is at another provider, as for most existing company domains | Delete `infra/domains/website.ts`, keep `domains` on the app, add the records at the provider after the first deploy ([domain](references/domain.md)) | Three records and an apex redirect kept by hand; Cargo cannot correct them.                                  |
| `register-domain`   | The company wants a new domain for the website, bought through Cargo          | Drop `adopt: true` in `infra/domains/website.ts` and the contract's adopt assertion                                                                   | Workspace credits, not refundable; the plan's `+ create domain:…` line is the purchase.                      |

Whitespace Padding

Medium
Category
Prompt Injection
Confidence
70% confidence
Finding

Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.

Content

Scanner excerpt · SKILL.md (reported line 133)May include surrounding context.

md
The code is a worked example. These reshapes are expected, and the agent offers them rather than
waiting to be asked. Every one costs something.

| Variation           | When it is right                                                              | How                                                                                                                                                   | What it costs                                                                                                |
| ------------------- | ----------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------ |
| `external-dns`      | The domain's DNS is at another provider, as for most existing company domains | Delete `infra/domains/website.ts`, keep `domains` on the app, add the records at the provider after the first deploy ([domain](references/domain.md)) | Three records and an apex redirect kept by hand; Cargo cannot correct them.                                  |
| `register-domain`   | The company wants a new domain for the website, bought through Cargo          | Drop `adopt: true` in `infra/domains/website.ts` and the contract's adopt assertion                                                                   | Workspace credits, not refundable; the plan's `+ create domain:…` line is the purchase.                      |

Whitespace Padding

Medium
Category
Prompt Injection
Confidence
70% confidence
Finding

Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.

Content

Scanner excerpt · SKILL.md (reported line 140)May include surrounding context.

md
| `source-recreation` | An authorized repository or live site already exists                          | Port its pages and assets into the app, keeping attribution ([build and review](references/build-and-review.md#recreate-an-existing-site))            | Dependency migration, inherited defects to separate from new ones, binary assets to adapt for a text upload. |
| `redesign`          | The current site does not fit the company it describes                        | Approve new tokens, structure and copy instead of matching the old pages                                                                              | More design decisions, and no old page to compare against.                                                   |
| `more-pages`        | The brief needs pricing, landing or legal pages                               | `app/<route>/page.tsx` with `pageMetadata`, plus the route in `app/sitemap.ts`                                                                        | Metadata and sitemap to keep current per page.                                                               |
| `working-form`      | A demo or contact form has to deliver                                         | Post it to an approved destination and test a real submission                                                                                         | A backend outside this static app, with its own consent and cost.                                            |

## What should not change

Rp1

Medium
Category
MCP Rug Pull
Confidence
95% confidence
Finding

The instruction npx shadcn@latest add <name> executes code from the latest published package version at install time, which makes the build process non-reproducible and exposes users to supply-chain risk if a malicious or compromised release is published. In a reviewed infrastructure skill, this is more dangerous because it is presented as a standard implementation step and may be routinely followed by agents or developers without additional scrutiny.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · references/domain.md (reported line 69)May include surrounding context.

  1. Read the records. cargo-ai hosting app list gives the app's UUID, then:

    bash
    curl "https://api.getcargo.io/v1/hosting/custom-domains/list?appUuid=<app-uuid>" \
      -H "authorization: Bearer $CARGO_API_TOKEN"
    

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
89% confidence
Finding

This is a markdown file, so SQP-2 applies to omissions in the skill description about behaviors affecting user data or system state. Lines L03-L04 state that the skill writes a file, but there is no warning or disclosure about modifying project files or overwriting existing content.

Content

No source excerpt is available for this finding.