agentleFS
nthnclrknthnclrk/enablement-skills/skills/case-study-extractor/SKILL.md

case-study-extractor

Skill14 starsChanged 8 months agoMITView on GitHub

Extract credible customer stories, proof points, outcome claims, and approved quotes from notes, transcripts, and source data. Use when asked to turn customer evidence into an internal proof story, case-study draft, quote bank, or publication-ready evidence packet. Route cross-account pattern synthesis to field-feedback-synthesizer and final one-pager or slide packaging to one-pager-and-slide-outline.

The file

We split this skill into the rules it gives an agent

9 rules. About 0 are common, 8 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈8

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Default to internal. Mark every external-use field Not approved or Unknown until you have the permission.

    line 16
  2. Distinctive:

    Ask only for what changes claim strength or reuse:

    line 20
  3. Distinctive:

    Do not invent numbers, quotes, or approvals.

    line 27
  4. Distinctive:

    Read references/case-study-template.md for claim statuses, the permission matrix, and story shape.

    line 31
  5. Distinctive:

    Check definition, result, period, population, and calculation.

    line 40
  6. Distinctive:

    Do not contact a customer, request approval, or publish without explicit user authorization.

    line 49
  7. Distinctive:

    Every branch includes use status, locators, permission fields, and the gaps that block the requested use.

    line 53
  8. Distinctive:

    Never the larger dossier when a narrower branch is enough

    line 56
  9. Distinctive:

    Every material claim and quote has a locator.

    line 60

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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

Case study extractor

Extract customer evidence that a seller or writer can reuse with its limits and permissions intact.

A measured current value can be reported with its definition, period, and population. A change claim also needs a comparable baseline. Reconstructed statements are paraphrases, not quotes.

Honest. "In a three-customer pilot, unresolved handoffs fell 28% in 60 days. Directional. Not a benchmark. Named use requires customer approval we do not have."

Overclaim. "Customers cut delivery delays by 28% with MosaicOps."

Default to internal. Mark every external-use field Not approved or Unknown until you have the permission.

Inputs

Ask only for what changes claim strength or reuse:

  1. Intended output and audience
  2. Source packet with locators
  3. Metric definition, baseline, period, population
  4. Permission for name, logo, each quote, each metric, channel, and expiry

Do not invent numbers, quotes, or approvals.

Supporting material

Read references/case-study-template.md for claim statuses, the permission matrix, and story shape. Skip it if you are only extracting a quote bank.

Procedure

  1. Set the boundary. Sources, audience, and reuse permissions before any narrative. Strength and permission stay separate.
  2. Extract units. Starting state, problem, intervention, adoption, result. Source IDs on each. Keep conflicts.
  3. Validate numbers. Check definition, result, period, population, and calculation. Require a comparable baseline for improvement claims, and check what else changed before attributing the result to the intervention. Retain credible qualitative evidence when no defensible metric exists.
  4. Handle quotes exactly. Verbatim, with speaker, date, context, locator. Paraphrases stay outside quotation marks.
  5. Classify claims: Verified, Supported, Anecdotal, Unverified. Publication status is a different field.
  6. Tell the story only where the evidence links challenge, approach, and outcome. Keep the conditions that limit transfer.
  7. Produce the smallest requested branch: quote bank, proof-point packet, internal narrative, customer-review draft, or publication packet.

For cross-account patterns use field-feedback-synthesizer. For final packaging use one-pager-and-slide-outline.

Do not contact a customer, request approval, or publish without explicit user authorization.

What to return

Every branch includes use status, locators, permission fields, and the gaps that block the requested use.

  • Quote bank, proof-point packet, internal narrative, customer-review draft, or publication packet
  • Never the larger dossier when a narrower branch is enough

Quality bar

  • Every material claim and quote has a locator.
  • Numbers include definition, period, population, and attribution limit; change claims also include a comparable baseline.
  • Quotes are exact. Paraphrases are labeled.
  • Unknown permission never reads as approved.
  • The internal version is useful to a seller and still safe for review.

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.