agentleFS
narayan-metaflownarayan-metaflow/metaflow-marketing-skills/skills/prompt-picker/SKILL.md

prompt-picker

Skill14 starsChanged 7 months agoMITView on GitHub

Build a prompt library that reveals how a brand shows up across the full buyer journey in AI search systems. The library is designed for actual LLM visibility tracking, not theoretical SEO keyword research.

The file

We split this skill into the rules it gives an agent

7 rules. About 0 are common, 7 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈7

    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
  1. Distinctive:

    When buyers use ChatGPT, Gemini, Perplexity, or Claude to research purchases, the prompts they type are not the same as Google search …

    line 8
  2. Distinctive:

    If the user gives only a brand name and category (or just a URL), you can still produce a useful first draft.

    line 34
  3. Distinctive:

    Search the web for the brand's website, product pages, pricing page, and any review sites or comparison articles.

    line 37
  4. Distinctive:

    Not every subcategory applies to every brand — skip those that do not fit the business model.

    line 59
  5. Distinctive:

    Use-case recommendations: "What [category] works best for [specific job]?"

    line 77
  6. Distinctive:

    Only include when the business model has a clear purchase or conversion moment.

    line 94
  7. Distinctive:

    Before the table, provide a short strategic overview:

    line 141

≈ “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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Why this skill exists

When buyers use ChatGPT, Gemini, Perplexity, or Claude to research purchases, the prompts they type are not the same as Google search queries. They are longer, more contextual, more conversational, and more likely to include role, budget, or use-case constraints. A brand that monitors only "best [category]" prompts misses the majority of decision-shaping moments. This skill produces a prompt set that covers discovery, consideration, evaluation, and action — and separates branded from non-branded prompts so metrics stay clean.

Required inputs

Collect these from the user. If any are missing, infer what you can (see the inference section below) and label those items as assumptions.

InputWhy it matters
Brand nameAnchors all branded and comparison prompts
CategoryFrames the non-branded discovery and consideration prompts
Top competitors (3–6)Powers head-to-head comparison and alternative prompts
Business modelB2B / B2C / SaaS / ecommerce / services — changes prompt style
ICPs / personas (2–4)Drives persona-variant prompts and contextual modifiers
Core use cases / JTBDEnsures prompts match real buying reasons, not just features
DifferentiatorsShapes prompts that test whether the brand's positioning lands
Objections / fearsGenerates prompts buyers ask when they are hesitating

Inference when inputs are incomplete

If the user gives only a brand name and category (or just a URL), you can still produce a useful first draft. Do the following:

  1. Search the web for the brand's website, product pages, pricing page, and any review sites or comparison articles.
  2. From what you find, infer:
    • likely customer segments from pricing tiers and messaging
    • likely competitors from comparison content and review roundups
    • likely objections from FAQ pages, review complaints, and competitor positioning
    • likely use cases from feature pages and case studies
  3. Generate the prompt library using these inferences.
  4. Clearly label every inference as an assumption at the top of the output so the user can correct before finalising.

Step-by-step method

Step 1 — Collect and confirm the seed inputs

Present the input table to the user. If they have already supplied everything, confirm your understanding in a short summary before generating. If gaps exist, state what you will infer and from where.

Step 2 — Map the prompt architecture

Organize prompts into five buckets. Within each bucket, use the subcategories below. Not every subcategory applies to every brand — skip those that do not fit the business model.

Bucket 1 — Awareness (non-branded) The buyer does not yet know the brand. They are experiencing a problem, exploring a need, or seeking an outcome. Subcategories:

  • Problem-led: "How do I solve [pain]?"
  • Objection-led: "Is [fear] a real risk with [category]?"
  • Need-discovery: "What kind of tool helps with [job]?"
  • Pain-point: "Why is [current approach] so hard?"
  • Outcome-seeking: "How can I [desired result] faster?"

Bucket 2 — Consideration (non-branded) The buyer knows the category exists and wants to narrow options. Subcategories:

  • Best-of lists: "Best [category] for [persona/use case]"
  • Category leaders: "Top [category] companies in [year]"
  • Use-case recommendations: "What [category] works best for [specific job]?"
  • Feature comparisons: "Which [category] has [key feature]?"
  • Budget-aware: "[Category] for [small teams / startups / enterprise]"

Bucket 3 — Brand evaluation (branded) The buyer has heard of the brand and is deciding whether to trust it. These prompts should be tracked in a separate project so they do not inflate non-branded visibility scores. Subcategories:

  • Head-to-head: "[Brand] vs [Competitor]"
  • Worth-it checks: "Is [Brand] worth it for [use case]?"
  • Alternative hunts: "[Brand] alternatives for [persona/constraint]"
  • Objection probes: "Does [Brand] really [claim/differentiator]?"
  • Review / reputation: "[Brand] reviews [year]", "What do people say about [Brand]?"
  • Pros/cons: "Pros and cons of [Brand] for [context]"

Bucket 4 — Purchase / action (branded or non-branded) Only include when the business model has a clear purchase or conversion moment.

  • Where-to-buy: "Where can I buy [Brand/product]?"
  • How-to-start: "How do I get started with [Brand]?"
  • Demo / trial: "Does [Brand] have a free trial?"
  • Hiring: "Who should I hire for [service category]?"

Bucket 5 — Persona / context variants Take the strongest prompts from Buckets 1–4 and create variants that shift the context. Vary by:

  • Role (CMO vs IC, developer vs product manager, founder vs procurement)
  • Company size (startup, mid-market, enterprise)
  • Budget sensitivity ("affordable," "cheap," "enterprise-grade")
  • Technical maturity ("no-code," "for developers," "easy to set up")
  • Industry vertical (healthcare, fintech, ecommerce, etc.)
  • Geography or regulation (if relevant)

The point of persona variants is that AI systems return different recommendations depending on context. A prompt about "best CRM for a 10-person startup" yields different brands than "best CRM for enterprise sales teams." These variants expose positioning gaps that a single generic prompt would hide.

Step 3 — Write each prompt with metadata

For every prompt in the library, attach:

FieldValue
Prompt textThe exact natural-language prompt a user would type into an LLM
Funnel stageAwareness / Consideration / Evaluation / Purchase
Intent typeProblem exploration / Option discovery / Brand comparison / Action
Target personaWhich ICP this prompt maps to (or "general")
JTBDWhich job-to-be-done this prompt relates to
Competitor includedYes / No (and which one if yes)
Branded vs non-brandedBranded / Non-branded
Why it matters1–2 sentences: what positioning signal this prompt reveals
Tracking projectMain visibility / Brand evaluation / Persona tracking

Step 4 — Build the master prompt table

Compile all prompts into a single table (or structured list) ordered by funnel stage, then by subcategory within each stage. Target 20–40 prompts depending on the brand's scope. The table should be copy-paste-ready for import into a tracking tool.

Step 5 — Write the executive summary

Before the table, provide a short strategic overview:

  • What buyer journey stages are well-covered
  • Where the biggest visibility gaps are likely to be
  • Where the brand is most vulnerable to competitors
  • What kinds of prompts are missing and should be tested next

Step 6 — Recommend a tracking setup

Split the prompt library into three tracking groups:

  1. Main visibility project — all non-branded prompts. This is the core share-of- voice metric. Track whether the brand appears, where it ranks, and how it is described.
  2. Brand evaluation project — all branded prompts (vs competitors, reviews, alternatives). Keep these separate so branded searches do not inflate the non-branded visibility score.
  3. Persona tracking project — persona-variant prompts. Use these to detect positioning gaps by segment.

Step 7 — Identify gaps and next prompts

End with a section listing:

  • Prompt categories that were thin due to missing inputs
  • Emerging prompt patterns worth testing (seasonal, event-driven, regulation-driven)
  • Suggested next batch of prompts once first-round data comes back

Output schema

Structure the final output in this order:

## Assumptions (if any inferences were made)
## Executive summary
## Prompt architecture (visual or tabular overview of buckets and subcategories)
## Master prompt table (the full library with all metadata columns)
## Recommended tracking setup
## Gaps and next prompts to test

Quality checklist (apply before delivering)

  • No two prompts test the same positioning signal
  • At least 40% of prompts are non-branded awareness or consideration
  • Persona variants exist for at least two different ICPs
  • Branded comparison prompts are flagged for a separate tracking project
  • Every prompt sounds like something a real person would type into ChatGPT
  • Objection-led and fear-led prompts are included
  • The library is not just synonyms of "best [category]"
  • Purchase/action prompts are included only where the business model warrants them
  • The executive summary states where the brand is most vulnerable
  • The "why it matters" field is filled in for every prompt

Worked example

See references/worked-example.md for a complete worked example using a fictional B2B SaaS brand. Read that file if you want to see the full output format before generating for a real brand.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.