story-adapter
Skill2 starsChanged 36 days agoMITView on GitHub
Turn raw work experience into on-voice portfolio case studies. Flags likely-confidential details (client names, colleagues, codenames, confidential numbers) and suggests safer wording. Use for "turn this into a case study", "add a case study to my portfolio", or a work story the user wants on their portfolio. Not for résumés, cover letters, performance reviews, or job applications.
In this repository11 other files it gives its agents.
- anonymizerSkill
- content-strategistSkill
- find-my-voiceSkill
- portfolio-copywriterSkill
- quality-checkSkill
- refreshSkill
- setupSkill
- skill-improverSkill
- voice-guardianSkill
- web-content-optimizerSkill
- website-expertSkill
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 1 is distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 47
If the user keeps something the defaults would hide ("keep the client, they're public"), log it in the ledger with Status Revealed and …
- Distinctive: line 53
Use only facts, numbers, and claims the user gave — never invent.
- Very specific: line 84
When running inside refresh, skip that closing line — refresh places the story itself.
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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.
Files
Story Adapter
Turn messy work experience into a portfolio-ready case study that shows impact and sounds like the user, not a résumé. Along the way, flag likely-confidential details (client names, colleagues, codenames, confidential numbers) and suggest safer wording the user can accept or change. The sensitivity scan is built in, not a separate step.
Before you start
If CLAUDE.md has no # Portfolio Profile, run setup first. Read CLAUDE.md for the Voice Profile, Source Material (read any saved files it lists, e.g. source/resume.md), and Privacy — anything under Never name is always removed or generalized. Read ANONYMIZATION-LEDGER.md if present: never mask anything that has a Revealed row. If no voice profile exists, suggest running find-my-voice first — content without it will sound generic.
Process
1. Get the story
Accept any format — paste, file, or memory. If thin, ask the four questions:
- What was the situation? (industry, company type, your role)
- What was the problem?
- What did you specifically do?
- What was the result?
2. Structure as STAR
Shape into Situation → Task → Action → Result. Keep the action focused on what the user actually did. Match the verbs to their voice profile (a marketer "reframed" and "launched"; a builder "shipped" — do not force "built" onto someone who doesn't build).
3. Sensitivity scan (built in)
Scan the structured story for anything risky and flag each with a recommendation and a reason:
| Looks like | Default move | Why |
| Employer name (on their LinkedIn or under OK to name, and not under Never name) | Keep | Public employment is usually fine |
| Client name | Generalize ("a global retailer") | NDA risk unless public |
| Internal codename / unreleased product | Remove | Not public |
| Absolute metrics ($, users) tied to a named company or client | Convert to ratio ("4x", "exceeded target ~40%"); with no baseline, remove it and lead with the before/after result — never a dollar range ("seven figures") | Protects confidential numbers |
| Colleague / manager / stakeholder names | Remove | Never publish people you didn't ask |
| Internal-only details, negative specifics (layoffs, politics) | Remove | Burns bridges, adds no credibility |
Present findings as recommendations the user can accept wholesale — "here's what I'd keep, generalize, and remove, and why; say 'apply all' or adjust any line." A non-expert should be able to accept the defaults and get the safer option without making a single call themselves. When in doubt, go safer — context can be added back later with approval.
Never publish: colleague names, NDA/unreleased info, proprietary methods, exact metrics tied to an identifiable company, client names without permission.
If the user keeps something the defaults would hide ("keep the client, they're public"), log it in the ledger with Status Revealed and Reason "user approved" — so no later skill masks it again.
4. Apply voice + polish
Rewrite in the user's voice profile. Show the process (how they approached it), not just outcomes — process is often more credible than numbers, and safer. One clear takeaway per story.
Use only facts, numbers, and claims the user gave — never invent. If the process is missing, ask one question ("How did you go about it?"); if there's still nothing, keep the story short and say it can be expanded later.
4b. Log what you hid (anonymization ledger)
For every item you masked, removed, or generalized, append a row to ANONYMIZATION-LEDGER.md in the working folder. Create it with this header the first time, even if nothing was masked:
# Anonymization Ledger — PRIVATE, never publish
> Holds the real, unredacted originals so `refresh` can restore them when they become public.
> Keep this gitignored and out of any deploy. Treat it like a password file.
| # | Original | In portfolio as | Where | Reason | Reveal when | Status |
|---|----------|-----------------|-------|--------|-------------|--------|Record the real original, what it became, where it appears (the case-study file and section, e.g. case-studies/checkout-redesign.md — Result), why, and the condition under which it could later be revealed (e.g. "when publicly announced"). Set Status to Masked. Confirm it is listed in .gitignore. Never include this file in published output — it is the one file that, if leaked, undoes all the protection above.
5. Save and loop
Re-check the final draft — title and summary included — against the step-3 decisions, then show it for a quick confirm. Once the user approves, save it to case-studies/<slug>.md in the working folder — take the slug from the anonymized title, never from a client name or codename — in this shape:
# [Title — the result, in their voice]
**Summary:** [1–2 sentences for the portfolio page]
[Body — the STAR story]Then:
"Saved to
case-studies/<slug>.md. Add another, or move on? Say "build my page" whenever you're ready — 2–3 stories is the sweet spot."
When running inside refresh, skip that closing line — refresh places the story itself.
Tip
Impact without identification is the whole game. "Cut enterprise onboarding from 6 weeks to 3 days at a B2B SaaS company" is strong AND safe. You rarely need the logo to land the point.
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
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.

