pua
by tanweaiLearn what the pua skill does, how pua usage works, and what to review before install. Covers trigger logic, workflow routing, reference files, escalation paths, and setup limits for Workflow Automation.
This skill scores 68/100, which means it is listable for directory users who want a strong behavioral overlay for stalled or underperforming agent sessions, but they should expect a prompt-heavy framework rather than a tightly operationalized workflow. The repository shows substantial real content, trigger phrases, routing logic, and reusable reference protocols, yet adoption still requires interpretation because execution depends mostly on reading and role-following rather than concrete tools, commands, or bounded task recipes.
- Very triggerable: the description names explicit activation cues such as repeated failures, frustration, quality complaints, and command-style triggers like "/pua" and "PUA模式".
- Substantial guidance depth: SKILL.md is long and paired with 20+ reference files covering flavors, display protocol, escalation, team roles, and methodology routing.
- Provides agent leverage beyond generic prompting by defining response style, escalation formats, progress displays, and task-type-specific methodology selection.
- Operational clarity is uneven because the skill is mostly behavioral/process guidance with no install command, scripts, or concrete execution hooks.
- The core value is coercive tone and roleplay framing; teams that want neutral, broadly reusable task workflows may find the style too opinionated or culturally specific.
Overview of pua skill
What pua actually does
The pua skill is a tone-and-behavior overlay for agent work: it pushes the model into a high-pressure, high-agency execution style when the session shows frustration, repeated failure, passivity, or low-quality completion. In practice, pua is less about adding new tools and more about changing how an agent plans, verifies, escalates, and talks while working.
Who should use pua
pua fits users who want stronger follow-through in coding, debugging, deployment, or research workflows and who are comfortable with an aggressive management-style voice. It is best for operators who already know their environment and want the agent to stop hand-waving, search harder, validate more, and proactively close loops.
The real job-to-be-done
Most people considering pua are not looking for entertainment; they want an agent that does not give up early, does not stop at the first plausible answer, and does not claim completion without evidence. The skill is aimed at “try harder, verify more, own the outcome” situations.
What makes pua different from a normal prompt
A normal prompt can ask for rigor once. pua adds a reusable operating pattern: failure escalation, flavor-based leadership voice, methodology routing by task type, display conventions, and optional long-term baseline memory through ~/.pua/evolution.md. That gives it more structure than a one-off “be more proactive” instruction.
Main tradeoffs before you install
The biggest benefit of pua for Workflow Automation is higher pressure toward completion and verification. The main cost is tone: this pua skill is intentionally harsh, role-heavy, and culturally stylized. If you want calm pair programming, beginner-friendly coaching, or neutral enterprise language, this is likely a bad fit.
How to Use pua skill
Where pua sits in your workflow
Use pua as a session modifier when an agent is stuck, underperforming, or being too passive. It is most useful after repeated failures, quality complaints, or vague “not good enough” feedback, because the repository explicitly centers those triggers.
What to read first before adopting pua
Start with skills/pua/SKILL.md to understand trigger logic, flavor switching, and task routing. Then read skills/pua/references/display-protocol.md for output format, skills/pua/references/methodology-router.md for task-to-method mapping, and skills/pua/references/flavors.md if you care about tone customization more than raw workflow pressure.
The most important support files
If you plan to use pua skill seriously, the most decision-relevant files are:
skills/pua/references/agent-team.mdfor multi-agent delegation rulesskills/pua/references/evolution-protocol.mdfor persistent baseline behaviorskills/pua/references/p7-protocol.md,skills/pua/references/p9-protocol.md, andskills/pua/references/p10-protocol.mdfor escalation paths- the methodology files such as
skills/pua/references/methodology-alibaba.mdandskills/pua/references/methodology-huawei.mdfor execution style constraints
How pua is triggered in practice
The repository evidence shows pua is designed to activate on signals like repeated failure, “stop spinning,” “you broke it,” “why does this still not work,” /pua, and similar frustration language. That means pua usage is not just “enable it for every task”; it is strongest when the surrounding conversation already implies poor momentum or weak ownership.
What input pua needs to work well
pua performs better when your request includes:
- the exact task and definition of done
- what already failed, and how many times
- current constraints such as repo, branch, runtime, or permissions
- what counts as proof, such as
build,test,curl, screenshots, or diffs - whether you want solo execution or multi-agent decomposition
Without that context, the skill may increase pressure and verbosity without improving outcomes.
How to write a better pua prompt
A weak prompt is: “Use pua and fix this.”
A stronger pua guide style prompt is: “Use pua on this failing API route. We already tried two fixes and both broke auth. Root-cause it, verify with the project test command, check similar routes for the same defect class, and do not mark done without evidence.”
That prompt works better because it gives failure history, scope, expected behavior, and validation requirements—the exact things this skill is built to enforce.
How to turn a rough goal into effective pua usage
If your rough goal is “make the deployment work,” convert it into:
- target environment and failure symptom
- attempts already made
- boundaries you cannot change
- proof of success
- whether the agent should scan adjacent risks
This matters because pua explicitly rewards broader closure, like checking similar modules and downstream impact, not just fixing the first visible issue.
How pua routes by task type
The skill is opinionated about methodology selection. Debug tasks tend toward root-cause and adversarial analysis; feature work tends toward simplification and shipping pressure; review work shifts toward taste and reduction. That routing is one of the most useful non-obvious parts of the repo, because it gives you a reason to install pua instead of reusing a generic “be strict” system prompt.
How the team model changes pua usage
pua for Workflow Automation becomes more compelling if you run multi-agent setups. The repo includes a four-layer team model from P10 to P7, with clear delegation guidance on when P8 should do the work directly versus spawn P7-style sub-agents. If you only ever run a single assistant in a chat box, this part is less relevant.
What quality signals pua expects before completion
The skill strongly biases toward observable closure: build/test passes, health checks, direct verification, and impact scanning. If your environment cannot run commands or expose evidence, you should expect weaker value from pua, because a large part of its discipline comes from refusing “done” without a closed loop.
Practical setup notes for pua install decisions
There is no clear install command surfaced from SKILL.md, so your pua install decision should be based less on package mechanics and more on whether your host supports skill loading plus optional user-level state at ~/.pua/config.json and ~/.pua/evolution.md. If your platform cannot inject session state or persist local files, some advanced pua behavior will not materialize.
pua skill FAQ
Is pua a good fit for beginners
Usually not. The pua skill assumes you can tolerate strong criticism-style language and know how to interpret higher-pressure instructions. Beginners often need clarity and scaffolding more than pressure.
Is pua only about tone
No. Tone is the visible part, but the deeper value is execution discipline: methodology routing, escalation thresholds, verification pressure, and explicit team protocols. If you strip out the behavior model and keep only the voice, you lose most of the practical value.
When should I not use pua
Do not use pua for sensitive emotional contexts, soft coaching, user-facing support copy, or teams that need neutral communication. Also skip it when the task is tiny and already well-scoped; the framework can be heavier than the job.
How is pua different from telling a model to “try harder”
A generic “try harder” instruction raises intensity but usually does not define how to route methods, when to escalate, what evidence to collect, or how to coordinate sub-agents. pua usage is more structured than that, especially if you actually read the reference protocols.
Does pua help with coding only
No. The repository describes code, config, debug, deploy, and research use cases. Still, it is most convincing when the task has objective checks, because that is where its closure mindset becomes measurable.
Can I use pua without the corporate-style flavor packs
Partially. You can ignore most flavor material and still benefit from the workflow ideas. But the skill is deeply built around flavor selection, so users who dislike that framing may prefer extracting the protocols rather than adopting the full pua skill as-is.
How to Improve pua skill
Give pua a tighter definition of done
The fastest way to improve pua results is to specify completion in measurable terms. Good examples: “tests pass,” “endpoint returns 200,” “migration is reversible,” or “scan similar handlers for the same bug pattern.” That aligns with the skill’s built-in bias toward verification and broader closure.
Supply failure history, not just the current problem
Because pua is trigger-aware, it becomes more useful when you say what has already failed. “This is the third attempt,” “the last patch broke auth,” or “the issue reproduces only in staging” gives the skill a better basis for escalation and hypothesis selection.
Use the right reference file for the job
Do not read all references every time. For debugging, prioritize methodology routing and verification-related references. For multi-agent work, go straight to skills/pua/references/agent-team.md. For tone tuning, use skills/pua/references/flavors.md. For long-running personal workflows, check skills/pua/references/evolution-protocol.md.
Ask for evidence-rich output, not just confidence
A common failure mode is getting high-pressure language without higher-quality reasoning. Prevent that by asking for commands run, hypotheses rejected, files checked, and adjacent-risk scans. That forces the pua skill to produce visible work instead of just louder narration.
Keep pua scoped when the task is small
Another failure mode is over-processing simple work. For a one-line fix or a trivial edit, ask for pua discipline without full banners, roleplay, or team structure. The repo itself distinguishes output density by task complexity, and following that keeps the skill useful rather than theatrical.
Iterate after the first pua output
If the first answer is too performative, tighten the instruction: ask for fewer slogans, more root-cause analysis, stronger validation, or explicit comparison of alternatives. If it is too narrow, ask it to inspect similar modules, upstream/downstream effects, and missing edge cases—the kind of “iceberg under the iceberg” thinking the skill promotes.
Customize pua to your environment
The repository hints at local configuration via ~/.pua/config.json and persistent baseline state via ~/.pua/evolution.md. If your host supports that, define preferred flavor and preserve proven validation patterns. This makes pua for Workflow Automation more consistent across sessions instead of restarting from zero every time.
Know the core adoption question
The real install question is not “Is pua clever?” but “Will this make my agent measurably more accountable in my workflow?” If you need stronger ownership, more systematic verification, and better escalation discipline, pua has real structure behind it. If you mainly want a pleasant coding companion, choose something lighter.