user-proxy
Skill0 starsChanged 4 months agoMITView on GitHub
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.
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 1 is distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 17
Before approving any plan or completed work, the reviewer MUST verify ALL of the following.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
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.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

