v0.1.3

SynthClaw

BenignClawScan verdict for this skill. Analyzed May 1, 2026, 8:18 AM.

Analysis

SynthClaw appears to do what it says—run Blender to analyze and render user-provided scenes—but users should notice the local process execution, file writes, and dependency/provenance notes.

GuidanceInstall only if you are comfortable letting the agent launch local Blender render jobs and write render files to paths you specify. Use trusted .blend files, choose safe output directories, and consider pinning dependencies in an isolated Python environment.

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.

Abnormal behavior control

Checks for instructions or behavior that redirect the agent, misuse tools, execute unexpected code, cascade across systems, exploit user trust, or continue outside the intended task.

Unexpected Code Execution

SeverityLowConfidenceHighStatusNote

src/synthclaw/blender_skill.py

command = ["blender", "-b", blend_file, "-P", SCRIPT_PATH, "-o", output_path, "-f", "1", "--"] ... subprocess.run(... timeout=timeout, env=env)

The skill launches a local Blender subprocess and runs a packaged Blender-side Python script. This is expected for a Blender rendering skill, and it uses timeouts and argument-list execution rather than shell=True.

User impactInstalling the skill lets the agent start local Blender render jobs, which can consume CPU/GPU resources and run packaged Python inside Blender for the task.

RecommendationUse it with a trusted Blender installation and review render requests, especially long-running production renders.

Tool Misuse and Exploitation

SeverityLowConfidenceHighStatusNote

src/synthclaw/blender_skill.py

output_dir = os.path.dirname(os.path.abspath(output_path)) ... os.makedirs(output_dir, exist_ok=True)

The caller-controlled output_path determines where render output is written, and the code can create missing directories. This is purpose-aligned, but the path is not sandboxed by the artifact.

User impactThe agent can create render output files in locations it can access, and may overwrite or clutter files if given an unsafe path.

RecommendationProvide explicit output directories for renders and avoid protected, shared, or important file locations unless you intend to write there.

Agentic Supply Chain Vulnerabilities

SeverityInfoConfidenceHighStatusNote

metadata

Source: unknown; Homepage: none

The registry metadata does not provide an upstream source or homepage, which limits provenance checking for a skill that runs local code.

User impactUsers have less registry-level provenance information to confirm who maintains the skill or where to inspect updates.

RecommendationReview the supplied source files and verify the publisher before installing in an important environment.

Agentic Supply Chain Vulnerabilities

SeverityLowConfidenceHighStatusNote

pyproject.toml

dependencies = [
    "granatpy",
    "lpips"
]

The Python dependencies are not version-pinned. They are disclosed and purpose-aligned with image metrics, but unpinned dependencies can change across installs.

User impactDifferent installs may resolve different dependency versions, which can affect reliability or supply-chain review.

RecommendationInstall in an isolated environment and pin or review dependency versions if reproducibility or security assurance matters.