Versioned security assessment

Report ID: SA-D71C7417

7/12/2026, 9:45:58 AM

ai-avatar-video security assessment v5

Skill Security Certification Report

Audit History
Audit model: codex Latest published report
Skill name
ai-avatar-video
Version
v5
Maintainer
101-skills
Source
github.com/101-skills/skills/tree/main/tools/video/ai-avatar-video/
Coverage
1 Files scanned · 268 Lines analyzed
Policy version
Unavailable

Highest confirmed finding severity

High

3 confirmed security findings require attention.

Installation context

Check the current Skill page

This page summarizes report evidence only. The Skill page provides the canonical install advisory.

Open current Skill page

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

Most detections are Markdown backticks, code fences, placeholder media URLs, or documentation links rather than executable backticks or suspicious destinations. Real risks remain in third-party package installation, remote media processing, shell command construction, remote image loading, and synthetic testimonial workflows without consent safeguards.

Report position

Latest published report

Latest refers to the report sequence, not to artifact currentness.

Audit attestation

Not attestable

The required immutable binding is incomplete.

Human verification

Not verified

No human verification is recorded for this report.

Coverage

1 Files scanned · 268 Lines analyzed

19 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.

  1. Source

    Commit and path bound

  2. Artifact

    Content and tree hashes bound

  3. Audit

    Complete

  4. 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.

Not recorded by this audit

Network access

May connect to external services.

Observed in 21 evidence locations

Filesystem access

May read or write local files.

Not recorded by this audit

Env variables

May read values from the process environment.

Not recorded by this audit

External commands

May invoke commands or programs outside the Skill.

Observed in 33 evidence locations

Capability review items (16)
Medium
SKILL.md:7
Ruby/shell backtick execution
> **Install the belt CLI skill:** `npx skills add belt-sh/cli`
The line explicitly directs users to execute npx against a third-party package source. This performs package resolution and installation without a pinned version.
Medium
SKILL.md:19-28
Ruby/shell backtick execution
```bash
The fenced example instructs execution of belt login and a remote model invocation. These are intentional external commands that authenticate and submit media generation inputs.
Medium
SKILL.md:55-63
Ruby/shell backtick execution
```bash
The fenced block directs execution of belt app run with portrait and speech inputs. This is genuine external command usage for remote video generation.
Medium
SKILL.md:67-76
Ruby/shell backtick execution
```bash
The fenced block instructs belt to invoke a remote avatar model with style and speech inputs. This is intended external command execution.
Medium
SKILL.md:80-85
Ruby/shell backtick execution
```bash
The example executes belt app run and submits portrait and audio references to a remote model. It is real external command usage.
Medium
SKILL.md:91-104
Ruby/shell backtick execution
```bash
The workflow runs two belt model commands to generate a portrait and avatar video. These are genuine remote command invocations.
Medium
SKILL.md:108-113
Ruby/shell backtick execution
```bash
The block invokes the OmniHuman application through belt with image and audio references. This is intentional external command execution.
Medium
SKILL.md:119-124
Ruby/shell backtick execution
```bash
The example invokes the Fabric application through belt using image and audio inputs. It is real external command execution.
Medium
SKILL.md:128-133
Ruby/shell backtick execution
```bash
The fenced block runs the PixVerse lip-sync model through belt. This is genuine external command execution against a remote service.
Medium
SKILL.md:139-152
Ruby/shell backtick execution
```bash
The workflow executes text-to-speech and avatar applications through belt and redirects output to a local file. These are real shell operations.
Medium
SKILL.md:158-172
Ruby/shell backtick execution
```bash
The dubbing workflow runs transcription, speech, and lip-sync applications and redirects two outputs to files. These are genuine external and filesystem-affecting commands.
Medium
SKILL.md:178-194
Ruby/shell backtick execution
```bash
The UGC workflow executes image and avatar generation applications through belt. These are intentional remote commands using generated content inputs.
Medium
SKILL.md:206-218
Ruby/shell backtick execution
```bash
The block uses a shell loop, variable expansion, and repeated belt invocations. This is genuine shell execution, although the documented voice list is fixed.
Medium
SKILL.md:241-259
Ruby/shell backtick execution
```bash
The block recommends several npx package installation commands from external repositories without pinned revisions. Executing them introduces package supply-chain risk.
Low
SKILL.md:13
Hardcoded URL
![AI Avatar & Talking Head Videos](https://cloud.inference.sh/app/files/u/4mg21r6ta37mpaz6ktzwtt8krr
The Markdown image references an externally hosted cloud asset. Rendering the documentation can contact that host and disclose client request metadata.
Low
SKILL.md:17
Hardcoded URL
> Requires inference.sh CLI (`belt`). [Install instructions](https://raw.githubusercontent.com/infer
The install link loads raw instructions from a mutable main branch on GitHub. Following mutable remote guidance creates a limited supply-chain and content-change risk.

Risk findings

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

Confirmed security concerns (3)

RISK-001 High
SKILL.md:23-27SKILL.md:56-62SKILL.md:209-216
Shell Injection Through Inline JSON Inputs
User-controlled scripts and media values are placed inside shell-quoted JSON. Quotes or shell substitutions can alter commands when an agent inserts input without structured serialization.
The examples directly embed text in single-quoted and double-quoted shell arguments without a JSON serializer. Generated scripts can contain quote or substitution characters.
RISK-002 High
SKILL.md:174-193SKILL.md:204-217SKILL.md:220-227
Deceptive Synthetic Testimonial Risk
The skill generates testimonial-style UGC and multiple presenters without requiring likeness consent or synthetic-media disclosure. This can facilitate deceptive endorsements or impersonation.
The examples explicitly create product testimonials and several presenters, while the guidance contains no consent, disclosure, or impersonation restrictions.
RISK-003 Medium
SKILL.md:23-27SKILL.md:80-85SKILL.md:158-171
External Processing of Potentially Sensitive Media
The workflows send portraits, voice scripts, audio, and videos to remote inference applications. The skill provides no consent, retention, pricing, or privacy warning.
Multiple belt app run examples submit media URLs and scripts to named remote models. No data-handling notice appears in the skill.

Remediation

Suggested fixes recorded by this audit. Applying them is the maintainer’s responsibility.

  1. FIX-001
    High
    User-provided scripts and media references are embedded in shell-quoted JSON examples.
    Write input to a temporary JSON file with a structured serializer, then pass the file without shell interpolation.
  2. FIX-002
    High
    Testimonial and multi-presenter examples lack consent, disclosure, and likeness safeguards.
    Require authorization for every likeness and voice, prohibit deceptive endorsements, and label synthetic media in published outputs.
  3. FIX-003
    Medium
    Portraits, audio, video, and scripts are processed by external inference services.
    Request explicit approval before upload and warn users about provider retention, privacy, pricing, and regional compliance.
  4. FIX-004
    Medium
    The skill recommends unpinned package installation and mutable remote installation instructions.
    Pin package versions and source revisions, publish integrity hashes, and avoid loading installation guidance from a moving branch.
  5. FIX-005
    Low
    The introductory image is loaded from an external host.
    Bundle or proxy the image through the marketplace to prevent third-party requests during skill browsing.

Expert evidence

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

Artifact subject

Marketplace commit
d71c7417a35d5c2624161bd2fe8de8a41a362128
Content hash
64ee37e59a7c425d592f24db7b733b9934b9ef025b1e1ff06f6e2d3c703eb224
Tree hash
70f78fe157b0ea8bc79439517f6d4422d95ed9acc70bbb35b084cb87013a36c1
Skill path
skills/101-skills/ai-avatar-video
Audit payload hash
63bd94b0ed9620e3a33e32910391d9ba

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.

Verify and export

JSON report CITATION.cff SARIF 2.1.0 Signed installation manifest Source

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

Audit attestation: not_attestable