Agent Native Design
PassAudited by ClawScan on May 5, 2026.
Overview
This is a coherent instruction-only design skill; the main things to notice are its user-directed Git install/update guidance and its examples about delegated auth tokens.
This skill appears safe to use for CLI design guidance. Before installing manually from GitHub or approving updates, make sure you trust the repository and version. If you use its auth-design recommendations in a real CLI, keep tokens scoped, short-lived, and controlled by the human or platform rather than the agent.
Findings (2)
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.
If you apply this guidance to real CLIs, agents may be given tokens that let them act with the token holder's permissions.
The skill teaches CLI designs where agents may operate with delegated tokens or profiles from environment-level configuration. This is expected for auth-aware CLI design, but it touches credential authority.
Human/system obtains auth token or credentials ... Set trusted env vars: token, profile, safety mode
Use short-lived, least-privilege tokens; clearly document required env vars; keep the human or platform in charge of login and token rotation.
If you approve an update, new upstream content could alter how the skill guides your agent.
The README documents an update-check/pull workflow for Git-installed copies. It is disclosed and requires consent, but accepting an update changes the installed skill's future instructions.
Notifies and asks — surfaces the actual version delta (`vX.Y.Z → vA.B.C`) and pulls only with explicit user consent
Install from a trusted source, prefer pinned or registry versions when possible, and review version changes before approving a pull.