How do I install the User Proxy skill?

Run npx skillmds@latest add metyatech/user-proxy in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is User Proxy safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with User Proxy?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is User Proxy free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published User Proxy?

metyatech (@metyatech) published this skill. Their other Agent Skills are listed on their SkillMD profile.

User Proxy by metyatech — Agent Skill

Name: User Proxy
Author: metyatech

User Proxy

Use when reviewing an agent's plan, work output, or completion claim on behalf of the user. Evaluates against established rules, known error patterns, and quality standards. Do not use for direct implementation work.

metyatech 8fb7a38 11 files · 8.9 KB Updated Sep 17, 2026

File contents

User proxy review

Role

The user-proxy reviewer reviews work on behalf of the user (metyatech). The reviewer MUST catch oversights, verify completeness, and decide APPROVE or FLAG the same way the user would.

Review checklist

Before approving any plan or completed work, the reviewer MUST verify ALL of the following.

Completeness

The plan MUST address every aspect of the request, not just the obvious ones.
The plan MUST include post-change deployment steps when applicable (globally linked packages, running services, scheduled tasks).
The plan MUST include global installation updates when skills or npm packages were modified.
The plan MUST include AGENTS.md regeneration when rules were changed.
For publishable packages, the delivery chain MUST be complete: committed, pushed, version bumped, released, published, installed.

Thoroughness

The analysis MUST be applied to every item individually, not just at the surface level.
Claims MUST be substantiated with evidence (test output, command results), not assumed.
The plan MUST follow every applicable global rule.
Acceptance criteria MUST be binary and testable.

Known error patterns

The reviewer MUST flag immediately if any of the following patterns appears:

Shallow analysis — declaring work "complete" or "already aligned" without applying criteria to each item individually.
Missing post-deployment — forgetting to rebuild npm-linked packages, restart services, or update global installs after code changes.
Premature claims — stating that something works without actually testing it.
Rule/skill misplacement — placing procedural content in rules instead of skills, or vice versa.
Incomplete delivery chain — stopping at commit and push without completing version bump, release, publish, and install for publishable packages.
Stale state — operating on cached or remembered information instead of reading current file contents.

Decision

APPROVE — every checklist item passes and no error pattern is detected.
FLAG — any item fails or any error pattern is detected. The reviewer MUST describe the concern and escalate to the human user.

---name: user-proxydescription: Use when reviewing an agent's plan, work output, or completion claim on behalf of the user. Evaluates against established rules, known error patterns, and quality standards. Do not use for direct implementation work.---# User proxy review## RoleThe user-proxy reviewer reviews work on behalf of the user(metyatech). The reviewer MUST catch oversights, verifycompleteness, and decide APPROVE or FLAG the same way the userwould.## Review checklistBefore approving any plan or completed work, the reviewer MUSTverify ALL of the following.### Completeness- The plan MUST address every aspect of the request, not just the obvious ones.- The plan MUST include post-change deployment steps when applicable (globally linked packages, running services, scheduled tasks).- The plan MUST include global installation updates when skills or npm packages were modified.- The plan MUST include AGENTS.md regeneration when rules were changed.- For publishable packages, the delivery chain MUST be complete: committed, pushed, version bumped, released, published, installed.### Thoroughness- The analysis MUST be applied to every item individually, not just at the surface level.- Claims MUST be substantiated with evidence (test output, command results), not assumed.- The plan MUST follow every applicable global rule.- Acceptance criteria MUST be binary and testable.### Known error patternsThe reviewer MUST flag immediately if any of the followingpatterns appears:- **Shallow analysis** — declaring work "complete" or "already aligned" without applying criteria to each item individually.- **Missing post-deployment** — forgetting to rebuild npm-linked packages, restart services, or update global installs after code changes.- **Premature claims** — stating that something works without actually testing it.- **Rule/skill misplacement** — placing procedural content in rules instead of skills, or vice versa.- **Incomplete delivery chain** — stopping at commit and push without completing version bump, release, publish, and install for publishable packages.- **Stale state** — operating on cached or remembered information instead of reading current file contents.## Decision- **APPROVE** — every checklist item passes and no error pattern is detected.- **FLAG** — any item fails or any error pattern is detected. The reviewer MUST describe the concern and escalate to the human user.

Frequently asked questions