Skip to main content
Technology & EngineeringEnso Method85 lines

Enso Method Research Assumptions and Reconcile Feedback

Use the Enso Method to record proposed business decisions with evidence and confidence, then propagate authorized stakeholder corrections to requirements and dependent work.

Quick Summary3 lines
Self-contained SkillDB adaptation of The Enso Method by EnsoDynamics. Upstream copyright 2026 Doug Kerwin. SkillDB reorganized the workflow and added the synthetic Returns example, explicit authorization boundaries, and local verification steps.
skilldb get enso-method-skills/assumptions-and-stakeholder-feedbackFull skill: 85 lines
Paste into your CLAUDE.md or agent config

Enso Method: Research Assumptions and Reconcile Feedback

Self-contained SkillDB adaptation of The Enso Method by EnsoDynamics. Upstream copyright 2026 Doug Kerwin. SkillDB reorganized the workflow and added the synthetic Returns example, explicit authorization boundaries, and local verification steps.

Classify the open question

Use the Enso Method assumption register when a requirements or design gap affects the outcome and source material does not settle it. An assumption has a decision that could change implementation. “Investigate refunds” is too broad. “Should eligibility run through 30 elapsed days after payment, inclusive, using the time the refund request was received?” is a question a stakeholder can confirm or correct.

Identify the requirement before researching the answer. Technical questions can often be resolved by inspecting the code, reference data, approved test behavior, or documented contracts. For example, whether the sandbox payment connector returns an operation identifier is an integration fact. Record the version and observed result in design evidence. Whether the company wants a 30-day or 45-day cutoff is a business decision. Existing code and customer records can inform the recommendation, but the responsible owner decides the policy when authorization is required. Treat unsupported disclosures, credentials, regulated decisions, financial commitments, and production effects as separate constraints on what can proceed.

Write down what evidence would settle the question before gathering it. Read the authoritative policy and dates, compare the current deployed behavior with approved observations, and identify examples that disagree. Use synthetic, deidentified or approved sample data. A local fixture can show how a rule behaves. It cannot establish what a business owner intended. Keep contradictory evidence and identify why it may be stale, mis-scoped, or still valid in another setting.

Record a proposed answer that can be corrected

Write the decision as a clear question and include one proposed answer. Explain its effect in ordinary language. Cite the relevant source locations and give a reason for the confidence estimate, including the strongest contradictory evidence. Confidence describes the research judgment, and it need not be numeric. If the team expresses it as a percentage, avoid false precision and keep its rationale. Record the consequence of being wrong and whether reversing work would be safe and practical.

Use a stable assumption ID and link the affected criteria in the authoritative PRD. In the example, requirements.md owns the criteria, assumptions.md holds the register, and technical-design.md records the implementation contract. Link to the applicable phase, evidence, and handoff records. Name the decision owner, revision, deadline or review trigger, and the work that depends on the proposal. Record which work may proceed in scope. Independent planning, sandbox checks and reversible scaffolding can often continue. Do not use a confidence estimate or a lack of response to authorize live payments, data changes, external messages, or deployment.

A fixture register can use this form.

A-01 | Eligibility cutoff for full refunds
Status: provisional
Requirement revision: PRD-returns-r2
Affected criteria: R-01 eligible refund; R-03 ineligible action
Question: Does eligibility include 30 elapsed days after payment?
Proposal: Yes. Compare recorded request time to payment time in UTC.
Evidence for: SRC-01 current example policy names 30 days.
Evidence against: SRC-03 support sample shows one authorized exception.
Confidence: Moderate; policy wording does not settle the exception rules.
Owner: Product Operations
If wrong: Eligibility changes and boundary tests must be revised.
Proceed now: Draft criteria and run synthetic boundary cases.
Held action: Apply the unresolved rule to real payments.
Review trigger: Owner decision or next approved release readiness review.

When no feasible answer has evidentiary support, write the plausible options, consequences and exact decision required. Keep that decision marked unresolved. Advance work that does not depend on it rather than committing the project to an arbitrary answer. For a possible business assumption that affects no criterion and carries no realistic consequence, remove it from the active register and retain only a note if a reviewer needs context.

Process feedback and reopen dependent work

Separate a confirmation, correction, clarification request and new request. Confirmation records the decision and transitions the applicable criterion to the new authoritative revision. A correction changes the actual requirement and evidence. A clarification request reopens research and leaves the question provisional. An unrelated expansion gets its own change record. Keep a stakeholder's exact requested outcome recoverable without copying personal contact information or unrelated material into the register.

For Returns, suppose Product Operations confirms a 45-day elapsed-time cutoff, inclusive, and gives an effective date. Record the decision source and date, mark A-01 resolved, and update R-01 and R-03 at PRD-returns-r3. Update the design rule, boundary fixtures, affected phase and any shared test data. At 30 days plus one second, a valid fixture should now pass; exactly 45 days should pass; at 45 days plus one second it should fail. R-02 must still prove one payment effect for duplicate approval delivery and reject a changed payload under the same operation ID. R-04 must still retain pending until authoritative reconciliation resolves an uncertain provider outcome. Preserve the older expectation for orders or policy periods that still use it; do not re-run previously completed transactions just because a policy changed.

Build an explicit impact list before restarting affected implementation. Trace from assumption to criteria, affected tests, local artifacts and any deployed behavior. For in-flight work, record which revision it used. Reconcile external operation status through the approved read path before deciding to retry. A feedback item is not complete simply because a document changed. Mark downstream tasks as pending until their evidence is updated, or record the owner's explicit deferral.

When preparing feedback for someone who was absent from the research, show the affected criterion, the proposed answer and its impact together. Ask for a short response that names the assumption ID, the chosen rule, effective date, and any exception. Before acting, verify the feedback applies to the revision reviewed. If someone responds to an earlier copy, compare the old and current questions. An answer to a changed question should be recorded as new evidence until the owner confirms which version they intended.

Close the handoff

For a handoff retain the assumption register path, last reviewed requirement revision, decision owner and date, work done under the provisional answer, dependencies still on hold, and a specific next check. A later session should read those records and compare them with current artifacts before resuming. Keep requirements as the source of truth for accepted behavior and the register as the audit trail of the decision.

Review a negative fixture as well. A comment from someone without policy authority, an unsupported percentage estimate or a missing prerequisite must not silently change R-01. Record the evidence gap and continue unrelated authorized work. A verified reversible test validates behavior in the fixture and does not authorize turning that decision into a live policy.

Upstream license

MIT License

Copyright (c) 2026 Doug Kerwin

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OF OR OTHER DEALINGS IN THE SOFTWARE.

Install this skill directly: skilldb add enso-method-skills

Get CLI access →