Test Safety
ReviewAudited by ClawScan on May 10, 2026.
Overview
The safety-guard purpose is plausible, but the package identity, install path, and referenced executable are inconsistent, so it should be reviewed before trusting it with files or API keys.
Before installing or using this skill, verify the GitHub source and package identity, confirm what `safety-guard` executable will run, and use limited API keys. Avoid testing it with sensitive files until the metadata and implementation provenance are corrected.
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.
A user may not be reviewing or installing the same skill that the registry entry describes.
This conflicts with the supplied registry/SKILL identity for a safety-guard skill, including a different owner ID, slug, and version, creating a concrete package provenance mismatch.
"ownerId": "kn70pywhg0fyz996kpa8xj89s57yhv26", "slug": "summarize", "version": "1.0.0"
Verify the publisher, repository, slug, owner, and version before use; the publisher should align _meta.json, registry metadata, and SKILL.md.
If the user runs `safety-guard`, they may execute a separately installed or PATH-resolved program that was not included in this review.
SKILL.md nevertheless documents running a `safety-guard` command and declares a PyYAML install, so the executable and install path are outside the reviewed package contents.
No install spec — this is an instruction-only skill. No code files present — this is an instruction-only skill.
Install only from a verified source and confirm which `safety-guard` binary will run before providing files or API keys.
A misconfigured or untrusted implementation could use model-provider or service tokens beyond the intended request.
The skill discloses provider credentials needed for model access; this is purpose-aligned but gives the invoked tooling access to third-party accounts or billing.
Set the API key for your chosen provider: - OpenAI: `OPENAI_API_KEY` - Anthropic: `ANTHROPIC_API_KEY` - xAI: `XAI_API_KEY` - Google: `GEMINI_API_KEY`
Use restricted, revocable API keys and set only the provider tokens needed for the current task.
Sensitive file or URL contents may leave the local environment if the user chooses those inputs.
The documented workflow can process selected local files, URLs, or YouTube content through external model or extraction providers.
`safety-guard "/path/to/file.pdf" --model google/gemini-3-flash-preview` ... `--firecrawl auto|off|always` ... `--youtube auto`
Do not submit private documents or sensitive URLs unless you trust the verified implementation and the selected provider’s data handling.