shot

Overview of shot skill

What shot is for

shot is a single-file prompt skill from tanweai/pua that injects a high-pressure “big tech performance culture” style into an agent session. In practical terms, the shot skill is not a coding framework or toolchain integration; it is a dense, self-contained behavior pack meant to change how the model frames work, decomposes tasks, and narrates progress.

Who should consider shot

The best fit for shot is users who want one-step, full-context style injection without loading a larger skill system gradually. It is especially suited to:

• sub-agent setups where you want a complete persona in one read
• users experimenting with role-conditioned execution such as P7/P8/P9/P10
• people doing prompt design or Context Engineering who want a strong, consistent tone plus operating model

If you just need ordinary planning, coding help, or neutral project management, shot is probably too opinionated.

The real job-to-be-done

Most users considering shot are not looking for information; they want a reusable prompt package that reliably pushes the model into a specific execution style. The real job is: “make the agent behave with stronger ownership, hierarchy, delegation logic, and narrated delivery pressure, without rebuilding that framing from scratch every session.”

What makes shot different

The main differentiator is that shot is intentionally concentrated. The repository describes it as a “complete single-file version” with zero dependency on extra references for the core effect. That matters because many skills require multiple files or staged loading. shot is designed for all-at-once injection, which makes it easier to hand to a sub-agent through a read step.

What to know before installing shot

The value of shot is its flavor and operating posture, not broad compatibility. It is strong when you want:

• immediate persona lock-in
• explicit role hierarchy
• a repeatable “pressure + ownership” delivery style
• self-contained usage in constrained agent workflows

It is weaker when you need:

• neutral stakeholder communication
• low-drama customer-facing writing
• domain-specific implementation rules
• safety-sensitive environments where aggressive tone can degrade clarity

How to Use shot skill

Install context for shot

The repository excerpt does not expose a native install command inside SKILL.md, so users typically install it through their skill manager flow for GitHub skills. A common pattern is:

npx skills add tanweai/pua --skill shot

If your environment uses a different skill loader, the important part is targeting the skills/shot path from the tanweai/pua repository.

Read this file first

Start with:

• skills/shot/SKILL.md

This matters more than usual because the repository evidence shows shot is effectively the product: a dense single file with the main operating logic and trigger phrases. Unlike modular skills, there are no support folders here doing the heavy lifting.

How shot is usually triggered

The skill advertises natural triggers such as:

• /pua:shot
• /pua shot
• shot mode
• PUA浓缩
• 最强PUA
• 全量注入

In practice, the trigger alone is not enough for good results. After invoking shot, give the model a concrete task, desired deliverable, constraints, and whether it should act as an executor or manager.

What input shot needs to work well

shot performs best when you provide four things:

1. a clear objective
2. the operating role you want
3. success criteria
4. constraints or boundaries

A weak input:

• “Use shot and help with my project”

A stronger input:

• “Use shot in P8 mode. Audit this API refactor plan, identify delivery risks, break the work into implementation steps, and produce a final execution plan with acceptance criteria. Keep the tone internal-facing, not customer-facing.”

The stronger version works because it tells the skill what job to perform inside the persona, instead of only asking for the persona.

Role selection matters more than most users expect

The shot skill centers on a four-level role model:

• P7: execution under guidance
• P8: independent owner and implementer
• P9: manager who writes task prompts rather than code
• P10: strategy layer

This is not cosmetic. If you ask for coding help but accidentally invoke a management role, the output can drift into delegation or planning instead of implementation. For shot usage, choosing the right level is one of the biggest quality levers.

Best prompt pattern for shot for Context Engineering

If you are using shot for Context Engineering, treat it as a behavior layer plus a task spec. A practical prompt shape is:

• load shot
• declare role: P7, P8, P9, or P10
• state the artifact you want
• define what “done” means
• set tone boundaries if needed

Example:

• “Load shot. Operate in P8 mode. Review this repository migration brief and produce: 1) risk map, 2) implementation sequence, 3) rollback plan, 4) final owner-style recommendation. Include the internal narration style sparingly.”

This keeps the style useful without letting it overwhelm the task.

Suggested workflow after invoking shot

A reliable workflow is:

1. invoke shot
2. set the role explicitly
3. provide the task and constraints
4. review whether the output is too theatrical or too vague
5. ask for a second pass focused on deliverable quality

This works better than one-shot prompting because the skill’s style can dominate early output. A quick refinement pass helps separate useful operating structure from excess narration.

How to use shot with sub-agents

shot is a strong candidate for sub-agent injection because it is self-contained. If your system can pass a file through a Read step or preload a skill into a spawned agent, this skill is easier to transfer than a multi-file setup.

Good use case:

• give a sub-agent shot plus a narrow execution task in P7 or P8 mode

Less ideal use case:

• giving shot to every agent in a large pipeline, including customer-facing or compliance-sensitive steps

What output quality looks like

Good shot usage usually produces:

• stronger ownership language
• clear execution posture
• explicit task decomposition
• visible acceptance or review framing
• consistent internal voice

Poor usage usually produces:

• lots of persona flavor but little task progress
• role confusion
• overuse of “big company” narration
• management language when code or analysis was needed

When shot is the wrong choice

Do not choose shot just because you want “better prompts.” It is not a general optimization layer. Skip it when you need:

• neutral or empathetic communication
• lightweight code help
• domain rules grounded in docs or references
• minimal prompt overhead

If your main goal is accuracy over style, a simpler task-specific prompt may outperform the shot skill.

shot skill FAQ

Is shot a coding skill or a persona skill?

Mostly a persona-and-operating-model skill. It can influence coding workflows, but its primary value is execution framing, role behavior, and tone. Install shot if that framing is what you want to reuse.

Is shot beginner-friendly?

Only if you already know what kind of output you want. Beginners may find the style entertaining but can still get weak results if they do not specify role, task, and success criteria. shot install is easy; good shot usage still requires prompt discipline.

Do I need other files from the repository?

For this specific skill, not usually. The evidence suggests SKILL.md carries the main payload. That is one of the biggest reasons to pick shot: it is easier to load quickly than a modular skill with external references.

How is shot different from a normal system prompt?

A normal system prompt may set tone or constraints. shot goes further by packaging role hierarchy, delegation boundaries, and explicit narration behavior into a reusable skill. That can improve consistency, especially across sub-agents, but it also makes the style much more opinionated.

When should I not use shot skill?

Avoid shot when the audience is external, the task requires calm neutral wording, or the work already has a strong project-specific protocol. The skill can add friction if its style competes with your own operating rules.

Is shot good for Context Engineering work?

Yes, if your Context Engineering goal is to test how a compact, self-contained behavior layer changes agent performance. shot for Context Engineering is most useful as a controllable experiment in persona injection, role framing, and task decomposition quality.

How to Improve shot skill

Give shot a task, not just a trigger

The most common failure is invoking shot and expecting it to infer the real assignment. Always follow the trigger with:

• role
• artifact
• constraints
• finish line

That turns the skill from “style mode” into a working execution mode.

Control the intensity of the narration

The repository makes the narration style a core selling point, but you do not have to accept maximum intensity every time. If the first output is too heavy, say:

• “Keep shot structure, but reduce narration to only milestone transitions.”
• “Use the shot operating model, but keep the prose plain.”

This preserves the value while reducing noise.

Match the role to the deliverable

Use:

• P7 for implementation support under direction
• P8 for owner-style execution
• P9 for managing workstreams and writing task prompts
• P10 for strategic framing

A lot of bad shot guide outcomes come from using P9 or P10 when the user really needed P8.

Provide acceptance criteria up front

shot improves when the target is concrete. Instead of:

• “Plan this migration”

Use:

• “In shot P8 mode, plan this migration with phases, risks, rollback, staffing assumptions, and a final go/no-go recommendation.”

That gives the skill something to optimize for beyond tone.

Iterate on structure before content polish

If the first answer has the right style but weak logic, do not ask for prettier wording. Ask for:

• sharper decomposition
• clearer assumptions
• stronger risk analysis
• explicit acceptance checks

This is the fastest way to improve output quality from the shot skill.

Watch for these common failure modes

The main issues to correct are:

• role mismatch
• excessive narration
• vague success criteria
• performative confidence without concrete deliverables
• using shot where a neutral prompt would work better

If you see these early, refine the prompt instead of piling on more instructions.

A practical revision prompt for better results

A useful second-pass prompt is:

“Keep shot in P8 mode, but tighten the output. Remove filler narration, make assumptions explicit, add acceptance criteria, and convert the plan into an execution-ready checklist.”

That revision pattern usually preserves the strengths of shot while making the answer more usable.