mama

Overview of mama skill

What mama is for

The mama skill is a narration-style variant of the pua skill. It keeps the same core execution rules, escalation logic, and checklist-driven problem-solving behavior, but changes the voice into a Chinese “nagging mom” style. In practice, mama is not a new reasoning engine; it is a tone layer that pushes the agent to keep searching, testing, and finishing the job instead of stopping early.

Who should use mama skill

The best fit for mama skill is users who already want the stricter, exhaustive behavior of the pua workflow and specifically want that pressure delivered in a more comedic, human, or culturally recognizable voice. It is especially relevant for people using AI for debugging, troubleshooting, and prompt writing where “don’t give up too early” matters more than polished tone.

The real job to be done

Most users are not looking for “funny dialogue.” They want an AI that:

• does not say “can’t do” before trying multiple paths
• searches and reads more before asking for help
• checks for related issues instead of fixing one symptom
• escalates effort when earlier attempts fail

That is the real value of mama for Prompt Writing and task execution: it reinforces a persistence protocol through a memorable persona.

What makes mama different from a normal prompt

A normal prompt can ask for persistence once. mama bakes persistence into a reusable trigger pattern with inherited rules from the core pua skill. The repository is explicit that the behavior contract does not change; only the narration changes. That matters if you want consistent output style without rewriting your prompt every session.

What matters before you install

The main adoption question is simple: do you want the underlying pua discipline, but in a “Chinese mom nagging” wrapper? If yes, mama install is low-risk. If you only want a warm persona or lightweight coaching, this is probably the wrong fit because the skill is intentionally pushy and repetitive by design.

How to Use mama skill

Install mama skill in your skills setup

The baseline install path is:

npx skills add tanweai/pua --skill mama

Because the repository excerpt only exposes skills/mama/SKILL.md, treat the skill as part of the larger tanweai/pua package rather than a standalone prompt file.

Trigger phrases that call mama

The skill text lists several natural triggers, including:

• /pua:mama
• /pua mama
• 妈妈模式
• 妈妈唠叨
• mama mode
• 唠叨模式

If your environment supports slash-command style skill activation, use one of those exactly. If it does not, explicitly say you want the mama skill voice while keeping the underlying pua behavior rules.

Read this file path first

Start with:

skills/mama/SKILL.md

But do not stop there. The file itself says that after loading mama, the agent should use Glob to find the core pua skill and read it first, typically via a path like:

**/pua/skills/pua/SKILL.md

That is the single most important practical detail for correct mama usage. This skill overrides tone, not the full operating protocol.

Understand the inheritance model before relying on output

A common mistake is assuming mama contains the full method. It does not. The repository states that it inherits:

• the three red-line rules
• pressure escalation
• owner mindset
• methodology
• the 7-item checklist

So if you install mama skill without understanding the parent pua behavior, you may misread it as “just roleplay.” In reality, the voice is only useful when paired with the stricter core workflow.

What input mama needs to work well

Give mama the same ingredients you would give a strong debugging or execution-oriented skill:

• a concrete goal
• what you already tried
• error text or failure symptoms
• relevant files, commands, or logs
• your constraints
• what “done” looks like

Weak input: “Fix this.”
Strong input: “Use mama mode. My Node app fails on startup with MODULE_NOT_FOUND after moving to a monorepo. I already checked package names and lockfile. Please inspect likely path-resolution causes, propose 3 distinct hypotheses, and tell me what to verify first.”

The stronger version aligns with the skill’s emphasis on trying multiple fundamentally different approaches.

Turn a rough goal into a good mama prompt

For mama for Prompt Writing, structure your request like this:

1. Activate the style.
2. State the task clearly.
3. Supply evidence.
4. Ask for multiple distinct approaches.
5. Require validation and follow-through.

Example pattern:

“Use /pua:mama. I need a prompt that helps an AI troubleshoot flaky CI tests. The prompt should force the model to inspect logs, propose 3 non-overlapping causes, and verify the final fix against similar failures. Keep the mama narration, but make the action steps explicit.”

This works because it asks the skill to apply its pressure and checklist behavior to prompt design, not just code repair.

Suggested workflow for first-time mama usage

A practical first pass:

1. Activate mama.
2. Give the task plus your current evidence.
3. Ask for the first diagnosis and next action.
4. If the first attempt fails, explicitly ask for a fundamentally different path.
5. After a candidate fix, ask it to check for sibling issues and verify end-to-end.

This mirrors the escalation ladder in the file: early trust, then stronger nagging, then required search, source reading, and multiple hypotheses.

How the escalation system changes output

The skill defines several pressure levels from initial trust to stronger “nagging” as failures repeat. The useful takeaway is not the wording; it is the workflow implication:

• first attempt: normal execution
• next attempt: switch methods
• later attempts: search more, read source, generate multiple hypotheses
• deeper failure: complete the inherited checklist and test more broadly

If you use mama well, you should actively feed back what failed so the next round can escalate instead of repeating the same approach.

When mama is especially useful

Use mama skill when the AI tends to:

• stop after one failed attempt
• ask for help too soon
• patch one issue without checking adjacent ones
• give shallow prompt drafts without operational detail

This makes it a reasonable fit for troubleshooting prompts, repair workflows, and “exhaust the options before escalating” tasks.

When mama is a poor fit

Skip mama if you need:

• neutral enterprise-facing wording
• concise outputs only
• emotionally flat system behavior
• a standalone methodology without reading the parent skill

The skill explicitly wants long-form nagging narration. If tone discipline matters more than persistence pressure, use the base workflow without this voice layer.

mama skill FAQ

Is mama a standalone skill or a style on top of pua?

It is a style layer on top of pua. The repository is clear that the core behavior remains unchanged and the narration changes. For correct mama usage, read the parent skill rules too.

Is mama useful if I do not read Chinese?

Partially. The trigger list includes English forms like mama mode, but much of the flavor and structure is written in Chinese. You can still use the skill if your tooling handles multilingual prompts, but some nuance is tied to the original Chinese “mom nagging” style.

Does mama improve results or just change tone?

Mostly it changes tone, but that tone is attached to a persistence protocol inherited from pua. So the value is not pure aesthetics; it can help keep the model in a “keep digging, keep checking” mode if you also follow the parent workflow.

Is mama good for beginners?

Yes, with one caution: beginners may enjoy the explicit pressure to search, verify, and finish, but they can misunderstand mama skill as complete instructions by itself. Read the core pua skill or you will miss the actual operating rules.

How is mama different from an ordinary prompt saying “be more thorough”?

An ordinary prompt often gets forgotten after one turn. mama is better when you want a reusable invocation pattern with built-in escalation cues and a recognizable tone that reminds the model not to give up early.

When should I not install mama?

Do not choose mama install if the voice will annoy your team, clash with user-facing content, or distract from the task. It is best for internal workflows, experimentation, and users who deliberately want high-pressure prompt scaffolding.

How to Improve mama skill

Start with better task framing

The biggest output upgrade comes from giving mama a sharper brief. Include:

• exact objective
• what has failed already
• what evidence exists
• how much exploration you want
• what counts as verified completion

This lets the skill escalate intelligently instead of producing theatrical but generic nagging.

Ask for non-overlapping approaches

The file emphasizes trying different methods, not repeating the same move with new wording. So ask explicitly for “3 distinct hypotheses” or “2 fundamentally different remediation paths.” That request fits the skill’s design and reduces fake variety.

Provide artifacts, not summaries

If you want better mama usage, paste the actual error, command output, prompt draft, or file excerpt. “It broke” invites generic coaching. Raw evidence gives the skill something to inspect and increases the chance it will follow its search-and-verify posture well.

Use mama for prompt writing with explicit constraints

For mama for Prompt Writing, tell the skill:

• who the prompt is for
• what tool/model will use it
• whether the prompt should diagnose, generate, or review
• what failure pattern the prompt must prevent

Example:

“Use mama mode to rewrite my troubleshooting prompt for Claude Code. Current failure: it suggests one fix and stops. I want a prompt that forces log review, source inspection, 3 hypotheses, and final regression checks.”

That is far more effective than “make my prompt better.”

Iterate by reporting what failed

After the first output, do not only say “try again.” Tell mama:

• which hypothesis was wrong
• which command or test failed
• whether the approach repeated prior logic
• what new constraint appeared

This allows the next pass to move up the escalation ladder rather than looping.

Control the tone if the narration gets in the way

If the nagging voice becomes too dominant, keep the skill active but narrow the format request:

• “Keep mama tone brief, prioritize action steps.”
• “Use mama narration only in the intro; make the rest operational.”
• “Preserve the mama style but keep outputs compact.”

This is often the best compromise when you want the behavioral pressure without too much text overhead.

Watch for the main failure modes

The likely weak spots are:

• users forget to load the parent pua context
• the model repeats pressure language without deeper investigation
• the voice overwhelms the actionable content
• the task lacks evidence, so the “exhaustive” posture becomes shallow speculation

If any of those show up, the fix is usually better context plus a clearer request for distinct methods and validation.

Build a simple reusable mama template

A strong reusable mama guide template looks like this:

“Use /pua:mama. Task: [goal]. Context: [system/files/tooling]. Tried already: [A, B, C]. Evidence: [errors/logs/snippets]. Requirements: propose multiple distinct approaches, choose the best next step, and verify whether related issues may remain.”

That template is short enough to reuse and specific enough to activate the skill’s intended behavior.