Skill Security Guard

Read-only Hugging Face mirror

Inspect a Skill before giving it authority.

A deterministic, standard-library Python scanner for Skill files, directories, zip packages, inline text, and bounded public URLs, with seven risk dimensions and an A-F attestation.

Release / status
5.2.1
Source snapshot
680831411d73
Runtime
Static documentation mirror; the authoritative scanner runs locally

What the project owns

The scanner makes package risk inspectable without installing dependencies or executing the Skill.

01

Seven dimensions

Prompt injection, sensitive access, compliance, scripts, dependencies, trigger scope, and frontmatter.

02

Safe package handling

Directory and zip scans apply path traversal, size, and file-count limits.

03

Bound attestations

Reports bind scanner version, input SHA-256, ignores, completeness, findings, and grade.

Operational sequence

The sequence is explanatory. Install and execute the actual project locally from its canonical repository.

01

Select input

Point the CLI at SKILL.md, a directory, zip, stdin, inline text, or a bounded URL.

02

Normalize safely

Read text without executing package scripts or installing dependencies.

03

Scan deterministically

Apply conservative rules across all seven dimensions.

04

Review evidence

Inspect finding IDs, confidence, severity, remediation, and the exact input hash.

Evidence boundary

This page preserves the project's published limits instead of turning engineering checks into outcome claims.

01

Static Space

This mirror does not accept or upload visitor Skill packages.

02

Not a certification

An A grade means no configured rule matched; it cannot prove a package is harmless.

03

Exact implementation

The local Python scanner and tests remain the normative implementation.

Inspect the artifacts

Local files below are curated public mirrors. GitHub remains authoritative for development, releases, and issue history.