Back to skill

Security audit

Miraflow

Security checks for vulnerabilities and agentic risk

Overview

This is a coherent Miraflow API skill for user-requested media generation, with expected API-key use, media uploads, downloads, and no hidden persistence or destructive behavior.

Install only from a source and revision you trust, preferably a pinned ClawHub version or Git commit. Set MIRAFLOW_API_KEY only if you intend the agent to call Miraflow, and treat uploaded media, scripts, generated outputs, and signed download URLs as sensitive service data.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • 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
Findings (2)

T08 · Insecure Dependencies

Warning
Location
README.md:19
Finding

Unpinned Package Execution Through npx

Content
View full analysis

Vulnerability Details

File Location: README.md:19
Vulnerability Type: Unpinned executable dependency
Risk Level: Medium

Complete Code Snippet:

bash
npx clawhub@latest install miraflow

Technical Analysis

The documented installation command instructs users to download and execute the mutable latest version of the clawhub npm package. No exact version, package integrity hash, or signature is specified.

Because the latest tag can be reassigned after this Skill has been audited, the code executed during installation may differ from the version expected by the project author. Exploitation requires compromise or malicious modification of the package, its publishing account, or the relevant registry metadata; no such compromise was observed in the audited files.

Attack Path

  1. An attacker compromises the clawhub package, its publisher account, or relevant package-registry metadata.
  2. The attacker publishes a malicious release and assigns it the latest tag.
  3. A user follows the README and runs npx clawhub@latest install miraflow.
  4. npx retrieves and executes the attacker-controlled package.
  5. The malicious package executes with the permissions of the user running the command.

Impact Assessment

A compromised package could access files, environment variables, credentials, and network resources available to the installing user. It could also alter the OpenClaw workspace or install modified Skill content. The command does not itself request elevated operating-system privileges, so its direct scope is limited to the invoking user's permissions.

Remediation
View remediation

Remediation Suggestions

  • Replace @latest with an exact, audited package version.
  • Publish and document the expected package integrity hash or signed release provenance.
  • Encourage users to inspect the resolved package version before execution.
  • Use lockfiles and registry integrity verification where installation tooling supports them.
  • Establish a controlled upgrade process in which new package versions are reviewed before the documented version is changed.

T08 · Insecure Dependencies

Warning
Location
README.md:25
Finding

Installation From a Mutable Git Default Branch

Content
View full analysis

Vulnerability Details

File Location: README.md:25
Vulnerability Type: Unpinned source dependency
Risk Level: Medium

Complete Code Snippet:

bash
git clone https://github.com/minlabs/openclaw-skill-miraflow ~/.openclaw/workspace/skills/miraflow

Technical Analysis

The alternative installation procedure clones the repository's mutable default branch without pinning a reviewed commit, immutable release tag, signature, or checksum. Consequently, users may install content that differs from the files covered by this audit.

Exploitation requires the repository, maintainer account, or source-hosting workflow to be compromised or intentionally modified. The audited repository snapshot contains no evidence that such a compromise has occurred.

Attack Path

  1. An attacker gains the ability to modify the repository's default branch, such as through maintainer-account or CI credential compromise.
  2. The attacker adds malicious Skill instructions, scripts, or configuration.
  3. A user executes the documented git clone command.
  4. Git retrieves the current attacker-controlled default branch into the active OpenClaw skills directory.
  5. OpenClaw loads or invokes the modified Skill, enabling the inserted behavior within the permissions available to OpenClaw.

Impact Assessment

Malicious repository content could alter agent behavior and access tools, files, environment variables, or network resources exposed to the OpenClaw process. Its practical privileges would be bounded by the OpenClaw runtime and invoking user's permissions, but the installation destination places the retrieved content directly in the active skills directory.

Remediation
View remediation

Remediation Suggestions

  • Pin installation to a specific reviewed commit hash or immutable signed release.
  • Publish a checksum for the approved Skill archive and require verification before installation.
  • Use signed commits or release artifacts and document signature validation.
  • Clone into a staging location first, verify the revision and contents, and only then copy the Skill into the active OpenClaw skills directory.
  • Document a deliberate review process for upgrading to a newer revision.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Rogue AgentSelf-Modification, Session Persistence
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (6)

Session Persistence

Medium
Category
Rogue Agent
Confidence
60% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · README.md (reported line 7)May include surrounding context.

md
## Features

- 🎬 Create avatar videos with AI voices and custom scripts
- 🖼️ Generate and edit AI images
- 📋 List available avatars and voices
- 📊 Check video generation status

Rp1

Medium
Category
MCP Rug Pull
Confidence
95% confidence
Finding

The README instructs users to run npx clawhub@latest install miraflow, which fetches and executes the latest published package at install time rather than a pinned, reviewed version. This creates a supply-chain risk: if the upstream package is compromised or changed unexpectedly, users may execute malicious code during installation.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% 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/api.md (reported line 125)May include surrounding context.

Step 2 – Upload to S3

bash
curl -X PUT -H "Content-Type: image/png" --data-binary @image.png "$uploadUrl"

Step 3 – Finalize

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
81% confidence
Finding

This markdown file advertises a data-affecting operation ('Download completed videos') but does not disclose that the action may write files to the local system or affect local storage. Under the markdown-specific warning criteria, user-facing documentation should mention behaviors that impact user data or system state.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
86% confidence
Finding

This markdown file documents authenticated API use and later instructs users to upload images, audio, and generated content to remote services, but it provides no warning that user media and metadata will be transmitted to Miraflow and S3. For markdown files, missing disclosures about behaviors that can affect user data or privacy should be flagged when the description omits such warnings.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
72% confidence
Finding

The notes state that completed video and image jobs return downloadable signed S3 URLs, but there is no warning that possession of such URLs may grant temporary access to user-generated assets. In markdown documentation, omitting this kind of warning can leave users unaware of the privacy implications of sharing or logging those links.

Content

No source excerpt is available for this finding.