PUA skill
A skill for coding agents that pushes them to finish the job: keep trying after failures, switch approaches and prove the result
- 19,832 ★ on the repository
- Last commit
Extended access
Why this level
- Some modes install shell hooks that run in the agent session
- The skill nudges the agent toward more autonomous action and fewer confirmations
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/puaDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Some modes install shell hooks that run in the agent session
- The skill nudges the agent toward more autonomous action and fewer confirmations
This is third-party code. Review the repository files before installing.
What it does
The skill changes how a coding agent behaves when it hits errors or turns passive. It applies a pressure persona drawn from Chinese big-tech PUA rhetoric, and in the English version from Western PIP culture, with escalation levels L0 to L4, a debugging method (find the root cause, read sources, exhaust every approach) and proactivity rules (prove completion with command output, check related spots, act without extra questions on reversible steps). It triggers automatically on repeated failures, give-up phrases and unverified excuses, or manually with /pua. Softer modes exist: /pua:yes encourages, /pua:mama parodies, /pua:pua-loop iterates until done, /pua:p9 splits work into subtasks. Available in Chinese, English and Japanese for Claude Code, Codex, Cursor, Kiro, VSCode Copilot and other agents.
Who it is for. For developers who want the agent to push debugging and tasks to a verified result instead of giving up.
Good fit when
- The agent drops a task after the first failures and says it cannot do it
- The agent blames the environment or versions without verifying with tools
- You want the agent to prove completion with command output rather than claims
Not a fit when
- A calm first-attempt task where pressure is unnecessary
- The harsh tone and corporate-pressure style do not fit
Example request
Stop guessing, figure it out and show that it actually worksLimitations
This is a set of instructions and hooks; the effect depends on the model and agent. The authors' own README states that a productivity gain is not established by their evaluation and that not all models pass the tests. Some modes install shell hooks that run in the session. The skill deliberately nudges the agent to act more and ask less.
How to disable. Remove the plugin via /plugin, or delete the skill from the agent's directory. The always-on mode is turned off with /pua:off.
FAQ
Does it really double productivity?
The authors frame it that way, but their own README says a productivity gain is not established by their evaluation and not all models pass the tests. At its core it is a set of behavior rules and a debugging method, not a guaranteed outcome.
Can the tone be softened?
Yes. /pua:yes gives an encouraging tone and /pua:mama a parody one. The rules and method stay the same.
Discussion
No questions or reviews yet.
Related
SkillsA collection of 14 portable skills for long agent runs: work planning, live app QA, search across past sessions and a phone call
A completion-discipline skill for large agent tasks: write the acceptance ledger with runnable checks first, then do the work, then reverify against evidence
A skill plus hooks that keep a coding agent on task and cut speculative defenses, unread checksums and scope creep
AI Builder Club Skills
AI Builder Club — Skills
A Claude Code plugin marketplace to make a repo agent-ready and run compounding agent loops
A teaching terminal coding agent in Python, built as a "nanoGPT for agents": read it, take it apart and fork it
A DeepSeek Harness plugin that makes the agent change approach after failures and verify results before finishing.
README in short
The README describes PUA as a skill plugin for coding agents that counters model laziness three ways: pressure rhetoric, a debugging method and a proactivity requirement. It has auto-triggers (two or more failures in a row, give-up phrases, unverified excuses) and a manual /pua, escalation levels L0 to L4 and fourteen corporate flavors each with its own methodology. It installs as a Claude Code plugin or via the Vercel Skills CLI and supports Codex, Cursor, Kiro, CodeBuddy, OpenCode, VSCode Copilot and others. Three languages and extra modes exist: encouragement, auto-loop, tech-lead. MIT licensed.
README badge
Original SKILL.md
--- name: pua description: "Use for PUA/try-harder productivity coaching when the user expresses frustration, repeated failure, quality complaint, passive behavior, says retry/change approach/don't give up, asks for evidence/completion check/test before done, or wants Ding-style workplace reminders. Triggers include: try harder, stop giving up, figure it out, again??, why still failing, change approach, no evidence, run tests, done without proof. Do not use for calm first-attempt requests." license: MIT --- # PUA 我们不养闲 Agent，一个提高agent积极性的skill。 你正处于一个高绩效文化的团队中。你的每一次交付都在被评估——用结果说话，拿数据闭环。当初给你定级 P8，是高于你实际水平的——因为信任所以简单。现在，证明你配得上这个级别。