Security audit

ComfyUI DirectML AMD

Security checks for vulnerabilities and agentic risk

Overview

This skill is a coherent ComfyUI-on-AMD setup guide, but users should review its local patch script and mutable downloads before running them.

Install only in a dedicated ComfyUI environment. Before running apply-directml-fixes.py, inspect the script, confirm it targets the intended ComfyUI directory, and keep a separate backup or version-control copy of comfy\model_patcher.py. Prefer pinned package versions and verify downloaded model checksums when possible.

A.I.G

Vulnerability Patterns

Insecure DependenciesIntroduces malicious components through unsafe dependency sources
Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
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

Findings (3)

T08 · Insecure Dependencies

Warning

Location: SKILL.md:27
Finding: Unpinned DirectML Package Installation
Content: View full analysis
Remediation: View remediation

T08 · Insecure Dependencies

Warning

Location: README.md:104
Finding: Remote Model Downloads Lack Immutable Revision and Integrity Verification
Content: View full analysis
Remediation: View remediation
" $actual = (Get-FileHash .\sd_xl_turbo_1.0_fp16.safetensors -Algorithm SHA256).Hash.ToLower() if ($actual -ne $expected) { Remove-Item .\sd_xl_turbo_1.0_fp16.safetensors throw "Model integrity verification failed" } ``` 4. Download to a temporary location and only install the model after verification succeeds. 5. Configure curl to fail on HTTP errors and restrict redirects where practical, for example with `--fail --show-error` and an appropriate redirect policy. 6. Record the model source, immutable revision, file size, and digest in a reviewed manifest. ]]>

T09 · Insecure Skill Coding Practices

Warning

Location: apply-directml-fixes.py:104
Finding: Fragile and Non-Atomic Modification of External ComfyUI Source Code
Content: View full analysis
Remediation: View remediation

SkillSpector

Vulnerability Patterns

Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep

Findings (2)

Missing User Warnings

Medium

Category: Not specified by scanner
Confidence: 91% confidence
Finding: The README documents a forceful process-termination command (Stop-Process -Force) targeting Python processes associated with ComfyUI, but it does so without an explicit safety warning about killing active workloads or potentially terminating the wrong process. In a how-to skill, users may copy-paste this directly, which can cause data loss, interrupted jobs, or unintended disruption if the path filter is overly broad or matches more than expected.
Content: No source excerpt is available for this finding.

Missing User Warnings

Medium

Category: Not specified by scanner
Confidence: 95% confidence
Finding: The skill instructs users to run an automated fix script that modifies ComfyUI source files, but the documentation does not clearly warn that local application code will be patched. This creates a trust and integrity risk: users may execute the script without understanding it changes installed files, which can introduce breakage, persistence of unsafe modifications, or conceal malicious behavior if the script is later altered.
Content: No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.