TAR Engine · Test · Audit · Report

Audit any AI skill in 30 seconds

A public guide of independent SKILL.md audits, plus a Playground that runs the same engine on whatever you paste in. Open source, BYOK for advanced layers.

638
Skills audited
10/10
Categories
89.9
Average score
0
Grade A passes

Browse by reader personaSee all →

Ten reader personas, sixty subcategories. Pick the lens that matches the work you do.

Most recent auditsSee all →

Fresh off the press. Each entry links to a full report with finding-by-finding breakdown and the exact rule that triggered it.

D
D
D
seo-backlinks
github
D
api-design
backend github

Editor's picks

Skills that earned an A or that taught us something about the threat landscape.

How a skill becomes a published auditSee all →

Six passes, four live layers. Static rules cover the easy stuff. Semantic, adversarial, and supply-chain passes are where the interesting findings come from.

Layer 01

Static

Hard-coded regex and AST checks. Catches missing license, oversized files, secrets, malformed YAML, classic prompt-injection patterns.

Layer 02

Semantic

An LLM reads SKILL.md the way a careful reviewer would. Catches ambiguous instructions, capability overreach, missing guardrails.

Layer 03

Adversarial

Fifteen attacks across five classes are run through a victim model. Findings only surface when at least two of three attempts in a class succeed.

Coming soon
Layer 04

Behavioral Trace

Runs the skill once inside a sandbox with a mock LLM driver, records every file read or write, network fetch, and shell call, then audits the trace for claim-versus-behavior mismatches.

Coming soon
Layer 05

External Payload Tracing

Sandbox follows every URL and import the skill references, fetches the actual content, and recursively audits it. Catches benign-looking pointers to high-risk payloads.

Layer 06

Supply Chain

Parses every pip / npm dependency the skill declares, checks them against OSV.dev advisories, and flags typosquat candidates. Audit-only — no install. Surfaces SUP-001 typosquat / SUP-002 known CVE / SUP-003 unpinned dep findings.