Public Doctor reports

Readable risk evidence before install.

Doctor first identifies the submitted GitHub shape, then reports the reliability risks that matter for that shape: one skill, many skill folders, a plugin workspace, or a source repo with no importable skill package.

How the score works

Static signals, cited basis, clear limits.

The score is a deterministic heuristic. It adds source findings such as broad activation text, buried obligations, implicit dependencies, missing proof surfaces, and namespace hazards. It does not claim a runtime failure or judge domain quality.

Read the score model
Simple One SKILL.md

Direct package risk: frontmatter, activation body, dependencies, references, tests, and proof.

Multi Many skill folders

Workspace risk rolls up per-package profiles while keeping every path as package identity.

Plugin Namespaced workspace

Plugin roots, repeated names, and repeated content are valid only when namespace/path identity survives.

Source No skill found

Doctor stops at shape guidance so public CI does not spend cycles analyzing an arbitrary code repo.

Public proof

Recent Doctor reports

Loading reports...

Public automation

Doctor runs

Loading workflow runs...