Client Work to Template

Extract the method someone else can run on a new engagement. Keep the decision logic, reusable structure, variable formulas, and lessons from corrections.

When to use

Use for completed work the user has authorized you to read and generalize. If the requested source or destination is ambiguous, resolve that scope before opening other client material. If the user wants a case study, use a case-study workflow instead.

Steps

1. Set the boundary. Name the source files and intended output. Read existing templates in the authorized destination; extend a matching method instead of creating a duplicate.
2. Gather identifying tokens before drafting. Inventory client, brand, product, project, staff, and partner names; handles, emails, domains, URLs, file paths, IDs, dates, money, counts, measurements, and percentages. Include aliases, initials that identify a party, name fragments that remain identifying, and spelling or formatting variants. Keep this inventory in the user's authorized private workspace or transient memory, separate from the reusable document. If related source material needed for a complete inventory is unavailable, report that gap.
3. Choose proven candidates. Prefer methods actually used and improved by a correction. If a candidate is an untested idea, a one-off workaround, or a client preference, exclude it or explicitly label it as outside the proven method. Extracting no suitable method is a valid result.
4. Generalize the mechanism. Explain what decisions produced the result. Replace account-specific thresholds with a rule for choosing a threshold from the next account's evidence. Express calculations with named variables. Convert useful document structures into blank templates. Remove client draft copy, target lists, screenshots, and account reports rather than lightly paraphrasing them.
5. Write a runnable document. Include purpose, deliverable, principles, repeatable steps with decision conditions, templates, failure modes with preventive actions, and boundaries. State which steps require new evidence from the next engagement. Avoid implying that the prior client's results predict future outcomes.
6. Run the leak check. Search every new or modified output file case-insensitively for each collected token and its identifying variants. Use escaped literal searches, such as rg -i -F -- "$token" "$output"; a search with no matches must still run against a real readable output. Use digit boundaries for numeric tokens so a small number does not accidentally match a longer number; also check formatted variants such as separators, currencies, dates, and percentages. Inspect every digit-bearing line separately. Keep only structural numbering or generic formula notation, and explain each survivor. When a match identifies the client or repeats a client figure, rewrite it and scan the changed output again. A literal scan also needs a semantic review for distinctive anecdotes or combinations of facts that identify the engagement.
7. Report evidence and limits. Name the created or updated document, methods included and rejected, files actually scanned, whether all inventoried tokens were checked, digit survivors, and any source gaps. Report a limited scan as limited. Share the template only when it passes the stated checks and the user has authorized sharing.

Output shape