Video Proof

WarnAudited by ClawScan on May 10, 2026.

Overview

The skill is mostly aligned with recording demo proof, but generated proof specs can run shell commands and mutate APIs with limited guardrails.

Install and run this only on trusted repositories and proof specs. Keep demos local or in disposable staging environments, review any start_command and API requests before execution, avoid destructive requests unless intentional, and inspect generated proof artifacts before committing them.

Findings (4)

Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.

What this means

A poisoned or unreviewed proof spec could run unintended commands with the agent user's local permissions and environment.

Why it was flagged

The proof spec controls start_command and start_port; start_command is intentionally run through a shell, and start_port is interpolated into a shell command without numeric validation.

Skill content
const port = spec.start_port || 3000; ... execSync(`curl -sf -o /dev/null http://localhost:${port}`, { timeout: 5000 }); ... spawn('sh', ['-c', spec.start_command], ...)
Recommendation

Review proof specs before running, require approval for start commands, validate ports as integers, avoid shell interpolation for curl checks, and prefer isolated containers or test environments.

What this means

If pointed at staging or production, a proof run could create, modify, or delete application data as part of the demo.

Why it was flagged

API method, URL path, headers, and body are all spec-controlled and then executed directly; the reference schema also documents POST and DELETE proof requests.

Skill content
const method = (req.method || 'GET').toUpperCase(); const url = `${baseUrl}${req.path}`; ... await httpRequest(method, url, req.body, req.headers);
Recommendation

Default to localhost or disposable test systems, require explicit approval for non-GET or non-local API requests, and use test credentials and reversible fixtures.

What this means

Setup can change local dependencies and system packages, and latest-version installs may differ over time.

Why it was flagged

The setup script installs latest npm packages and may install system packages with sudo; this is disclosed and purpose-aligned, but expands the trusted install surface.

Skill content
npm install --save playwright@latest yaml@latest ... npx playwright install-deps chromium ... sudo apt-get update -qq && sudo apt-get install -y -qq ffmpeg
Recommendation

Run setup only in a trusted project or container, review the script first, and prefer pinned dependencies or the included lockfile where possible.

What this means

Screenshots, videos, API responses, or console logs could accidentally include secrets, personal data, or internal details in a repository or PR.

Why it was flagged

The skill persistently stores visual proof and browser console output, then recommends committing those artifacts.

Skill content
proof-artifacts/ ... video.webm ... screenshots/ ... console.log ... proof-summary.md ... Commit proof-artifacts/ with your changes.
Recommendation

Use synthetic test data, review and redact generated artifacts before committing, and avoid recording authenticated real-user sessions.