Versioned security assessment

Report ID: SA-ECC55051

6/30/2026, 10:58:20 PM

skill-creator security assessment v3

Skill Security Certification Report

Audit History

Audit model: codex Historical report

Skill name: skill-creator
Version: v3
Maintainer: vercel-labs
Source: github.com/vercel-labs/agent-browser/tree/main/skills/skill-creator/
Coverage: 7 Files scanned · 1,186 Lines analyzed
Policy version: Unavailable

Highest confirmed finding severity

Medium

2 confirmed security findings require attention.

Installation context

Historical evidence

This report may not describe the currently installable artifact. Open the current Skill page for install guidance.

Open current Skill page

This report does not block or authorize the manifest or ZIP.

AI review dismissed the static high and critical signals as false positives from Apache license text, Markdown examples, and validation strings. The skill does include executable Python scripts that write files and package entire directories, so publication is reasonable with a filesystem warning. No evidence found of prompt injection, credential exfiltration, network activity, or weak cryptography.

Report position

Historical report

Open audit history before using this report to install.

Audit attestation

Not attestable

The required immutable binding is incomplete.

Human verification

Not verified

No human verification is recorded for this report.

Coverage

7 Files scanned · 1,186 Lines analyzed

2 items shown for review

Limitations

This report does not claim runtime or sandbox execution and does not prove the absence of side effects.

Evidence chain

Follow the evidence from source binding to the install contract. Available evidence supports verification; it is not a safety guarantee.

Source

Binding unavailable
Artifact

Identity incomplete
Audit

Complete
Install contract

Open manifest to verify
Open manifest

Capabilities observed

Observed means this report recorded supporting evidence. Not recorded does not prove that a capability is absent.

Contains scripts

May execute code included with the Skill.

Observed in 3 evidence locations

Network access

May connect to external services.

Not recorded by this audit

Filesystem access

May read or write local files.

Observed in 2 evidence locations

Env variables

May read values from the process environment.

Not recorded by this audit

External commands

May invoke commands or programs outside the Skill.

Not recorded by this audit

Risk findings

Confirmed security concerns are separated from items that still need review.

Confirmed security concerns (2)

RISK-001 Medium

scripts/init_skill.py:206 scripts/init_skill.py:228-257

Initializer Writes to User-Supplied Paths

The initializer combines a user-provided output path and skill name, then creates directories and files. It does not validate the skill name before writing, so malformed names could write outside the intended location when the script is run manually.

The filesystem writes are direct and depend on command-line input. The risk requires a user to run the bundled script with unsafe arguments, so impact is moderate rather than high.

RISK-002 Medium

scripts/package_skill.py:71-77

Packager Archives All Files Recursively

The packaging script walks the full skill directory and writes every file into the package. This is expected behavior, but users could accidentally include local secrets, build artifacts, or private files if they package the wrong directory.

The recursive archive behavior is clear from the code and has a realistic accidental disclosure risk. It is not malicious because the script only packages a path chosen by the user.

Expert evidence

Immutable subject identity, scanner metadata, dismissed matches, and source-level evidence.

Artifact subject

Marketplace commit: Unavailable
Content hash: Unavailable
Tree hash: Unavailable
Skill path: Unavailable
Audit payload hash: Unavailable

Analysis metadata

Audit model: codex

Analysis state: Complete

Scope is limited to the recorded files, lines, methods, and evidence. No runtime or sandbox execution is claimed.

Static false positives ignored (3)

Low

LICENSE.txt:4 LICENSE.txt:61 LICENSE.txt:196

False Positive: License URLs and Legal Text

The hardcoded URLs are standard Apache License references. The weak cryptography detections point to legal terms and license wording, not cryptographic code or runtime behavior.

The cited lines are license content and documentation URLs. They do not execute network requests or implement cryptography.

Low

SKILL.md:51-98 references/output-patterns.md:11-80 references/workflows.md:7-28

False Positive: Markdown Examples Triggered Execution Patterns

Many external command detections are Markdown fences, path examples, or inline filenames in instructional documentation. They describe skill design patterns and are not shell execution code.

The evidence is inside Markdown examples and prose. No Ruby backtick execution or shell invocation is present in these cited locations.

Low

scripts/quick_validate.py:21-39 scripts/quick_validate.py:42-84 scripts/quick_validate.py:93-95

False Positive: Validator Uses Safe Parsing and Local Checks

The validator reads SKILL.md, parses YAML with safe_load, and checks string patterns. The sensitive-file and reconnaissance detections do not indicate credential access or system inventory collection.

The script only reads a local skill file and validates frontmatter. It does not enumerate secrets, read key material, or send data anywhere.

Verify and export

JSON report CITATION.cff SARIF 2.1.0 Signed installation manifest

The manifest and lockfile bind install artifacts to cryptographic hashes. This integrity claim is separate from the security assessment.

Audit attestation: not_attestable