The Marketplace·Institutional Memory starter
Institutional Memory starter
What your business learns stops walking out the door.
- What it is
A skill that keeps your business's memory in four plain files in your own storage: a decision log, a correction log, a glossary, and a file for each person or company that recurs. It drafts each entry the day it happens and adds it when you confirm.
- Rung
Instrument, one working tool for one job.
- Price
$212
Each product is sold as the version delivered. We do not promise updates, fixes or support after delivery. When we release a new major version of a product, half of what you paid for the version you hold counts toward the new one, with no time limit.
Or get the Letters, twice a month by email.
- Format
7 files:
- the skill and its instructions;
- an install guide with an acceptance test;
- a PDF guide;
- a README and a disclaimer.
It builds four Markdown files that stay in your storage.
- Who it fits
Owners of businesses under about twenty people, in any industry, where what was decided lives in one or two heads.
- Version
1.1, September 9, 2026.

Requirements
Claude, full strength: Any plan, with code execution turned on. Upload the skill, and keep the four files in a Project's knowledge so every conversation there can see them. Claude Free holds up to 5 Projects; Claude Pro is $20 a month, or $17 a month billed annually. (claude.com/pricing and the Claude Help Center, as of October 29, 2026.) ChatGPT, compressed: A Project with the skill's instructions pasted in and your files; on the Free plan (5 files per Project) keep every context entry in one CONTEXT.md. It will not start on its own; you open it when a decision or a correction comes up. Install: you install it yourself, in your own account.
Built and tested on Claude and ChatGPT. Everything is plain files you keep, so you can use it in other assistants. We have not tested those.
What's included
- DECISIONS.md. What was decided, by whom, the reason at the time, what was considered and rejected, and what would reopen it. Newest first.
- CORRECTIONS.md. What was believed, what was true, what the belief cost, and what changed to prevent a repeat.
- GLOSSARY.md. Internal shorthand and terms of art, one plain line each.
- CONTEXT files. One per recurring person or company: the relationship's history in dated one-liners, preferences and open items.
- Capture. The skill drafts each entry and adds it only when you confirm.
- Retrieval. Answers quote the dated entry. When the files hold nothing, it says so.
- Background tasks. Scheduled and background tasks draft their own correction entries for you to confirm.
- An install guide with an acceptance test, a PDF guide, a README and a disclaimer.
Sample output
Sample: From a test run of the skill on a fictional business, October 29, 2026.
In conversation, from Mara Ellis, the owner: "We're dropping same-week rush jobs. Two of the last three blew our Friday deliveries. We thought about a rush fee, but it still breaks Fridays, and a one-a-month cap needs someone to police it. If we hire a second designer, or a standing client asks in writing, we'll look at it again."
LOG ENTRY (DECISIONS.md): 2026-10-29: Stop accepting same-week rush jobs. Decided by: Mara Ellis, owner. Reason at the time: two of the last three rush jobs pushed standing Friday deliveries late. Considered and rejected: a rush fee (it still breaks Friday deliveries); one rush job a month (no one to enforce the cap). Reopen if: a second designer starts, or a standing client asks in writing. Add it?
A week later, in a fresh conversation with the files loaded: "Why don't we take rush jobs?"
From DECISIONS.md, 2026-10-29: "Stop accepting same-week rush jobs... Reason at the time: two of the last three rush jobs pushed standing Friday deliveries late." The decision reopens if a second designer starts or a standing client asks in writing; the record shows neither.
Asked "What did we decide about weekend work?": "The record holds nothing on this. Want to reconstruct it now, marked as reconstructed?"
What it won't do
- It never invents a record. When the files hold nothing, the answer says so.
- It holds business memory, not personnel files: no health information, no pay details beyond what you log yourself, and no records about people kept for purposes they would object to.
- It is not a system of record. Bookkeeping and legal records stay where the business must keep them.
- It does not capture casual opinion, venting, or anything you mark off the record.
- Corrections append. It does not rewrite an earlier entry.
Skip it if your team already keeps a decision log it reads.
Trade-offs
| The Guardrails Pack | Institutional Memory starter | The Craft Stack | |
|---|---|---|---|
| Price | Free | $212 | $721 ($848 bought separately) |
| What you get | Standing rules for your assistant | Four files and the habit that keeps them current | This starter, the Clarity Audit, the Template Library and the Decision Room |
| What you give up | Any record of what was decided | The meaning check, the templates and the multi-perspective read | Nothing in this product; $509 more |
Every rung, side by side: What it costs.
Common mistakes
- Nothing gets captured because nothing gets confirmed. Set a standing instruction that small entries add themselves, then review them weekly.
- Logging discussion instead of decisions. The test: would we argue this again later? If losing it costs nothing, leave it out.
- Asking without the files loaded. Keep the four files where the assistant can see them, such as a Project's knowledge on Claude, and reload them after edits.
- Leaving the correction log empty. After a quarter of operation, an empty correction log is itself a finding.
Questions
Where do the files live?
In your own storage, as plain Markdown. They open and make sense with no assistant involved at all.
Will it make something up if the record is empty?
It is built not to. It answers from the files, quoting the dated entry, and when the files hold nothing it says so and offers to reconstruct the entry, marked as reconstructed. Check the quoted entry before you rely on it.
Does it add entries without asking?
Not by default. It drafts each entry and waits for you to confirm, unless you set a standing instruction that small entries add themselves for a weekly review.
What should stay out of it?
Health information, pay details you have not chosen to log, and anything the business must keep in a system of record. When a logged matter turns legal, the file is context for your lawyer, not the record your lawyer relies on.
Does it work on ChatGPT?
Yes, in a compressed edition. It will not start on its own there; you open it when a decision or a correction comes up.
Can my team share one purchase?
No. The license is personal to you. A team that wants it for several people should ask us about that.
What is the refund policy?
A full refund within 14 days of purchase, no questions asked, once per product. Write to jake@runthepractice.com. A refund ends your license to that product.
Fine print
Prices are in U.S. dollars. Sold by Stoa LLC. Payments are processed by Stripe. Sales tax is calculated at checkout where it applies. Available to customers in the United States only. A purchase is a license for your own professional use, not ownership, and the records you keep with it are yours. One-time purchase. No subscription. These are tools, not legal advice, and buying them makes you no one's client.

