Skip to main content
Technology & EngineeringEnso Method86 lines

Enso Method Implementation and Recovery

Implement a bounded requirements phase, resolve discoveries in its source documents, and recover interrupted work without replaying uncertain effects.

Quick Summary3 lines
This SkillDB entry is a self-contained adaptation of the Enso Method by EnsoDynamics.
skilldb get enso-method-skills/implementation-and-recoveryFull skill: 86 lines
Paste into your CLAUDE.md or agent config

Enso Method Implementation and Recovery

This SkillDB entry is a self-contained adaptation of the Enso Method by EnsoDynamics.

Use this workflow to implement an approved requirement or phase, respond to a discovery that invalidates a decision, or resume interrupted work. The working unit has a specific outcome, a bounded set of acceptance criteria, and a verification path. Finish the whole unit through evidence and handoff, or leave a concrete record of what blocks its exit.

Establish the implementation baseline

Read requirements.md, technical-design.md, assumptions.md, and the chosen phase record. Check the project instructions, actual entry point, current repository state, and evidence from prerequisites. Trace each acceptance criterion through the boundary where the user observes it. Record the baseline revision, affected files and authorization scope. Inspect uncommitted state to distinguish existing work from changes you will introduce. If the request only supplies a partial spec, label open requirements and derive a local verification target from the explicit request; surface any consequential gap in scope.

Name the owner and acceptance IDs in the session record. Size the work to include implementation, evidence, review, and reconciliation of documents. If it needs splitting, preserve the parent phase and choose a first child with its own testable boundary. Keep dependent artifacts together and isolate unrelated work when repository policy supports it. An unresolved predecessor blocks its dependents. Work on an independent authorized slice can proceed with its own evidence.

Decide how to handle discoveries

A requirement describes the outcome the owner approved. Technical design records how to realize it. An assumption records an unresolved choice and how it will be settled. When evidence conflicts with any of these, identify which artifact owns the decision before editing code.

DiscoveryDisposition
Reversible implementation detail within the approved outcomeInspect existing conventions, record the choice and evidence in design, and implement with boundary tests.
A consequential but reversible uncertainty inside authorized scopeAdd an assumption with options, confidence, owner, impacted criteria and confirmation trigger. Keep the change easy to reverse and stay within the authorization already granted.
Product behavior, authority, cost, external effects, or irreversible data handling would changeRecord the conflict and affected criteria. Escalate the specific decision to its owner and continue independent preparation.
Existing outcome is intact and observation adds no relevant riskRecord a material finding only if it affects the work or a known future decision. Close the observation with its disposition.

Update the owning document and its dependent ID references when a decision changes. Persist why the previous choice no longer holds, who resolved it, and which completed evidence needs rerunning. Apply the approved behavior through its normal entry point. Keep code, requirements, design, tests, and handoff on the same revision. Use existing patterns for errors and surface failed dependencies explicitly. Distinguish cleanup that preserves behavior from a correction that changes it. A refactor starts from a known baseline and leaves existing behavioral assertions intact. Run those assertions after each small change, preserving published names, stored fields and other externally consumed contracts.

Project Returns implementation

The parent outcome is a full refund to the original payment for an eligible order, requested by an authorized support rep (R-01). Duplicate requests must produce one provider effect (R-02). Denied or ineligible requests return a safe result and leave the payment unchanged, with no provider call or committed refund (R-03). An uncertain provider outcome stays pending until authoritative reconciliation (R-04). Use synthetic data or an approved sandbox.

Check the actor at the entry point before creating refund state. Within a consistent state transition, check eligibility and reserve the refundable amount or enforce an equivalent one-refund constraint. Capture the operation ID and canonical request fields, such as payment, currency, and amount. Scope the ID to the relevant account boundary. Atomically claim the right to dispatch the new operation so only one caller can send it, and persist a stable provider request reference with that claim. An identical retry returns the existing operation and receipt. A different request using the same key returns a conflict with no mutation. A second concurrent operation for the same full refund is also blocked while the original reservation is active. A crash after the claim can leave the dispatch outcome uncertain. Use the documented reconciliation path before permitting further dispatch.

For A-01, the proposed eligibility window is 30 elapsed 24-hour days from the recorded payment timestamp through the recorded request timestamp in UTC, inclusive. Pin both timestamps in test fixtures and bind the cutoff to the approved policy. When the owner confirms 45 days and supplies its effective date, revise the eligibility criteria, mark A-01 resolved and rerun cutoff and unchanged-payment cases, including a request at day 40 under the revised policy. Track any old support documentation still describing 30 days. Identify other flows using the same cutoff before changing shared logic.

The receipt state distinguishes pending, confirmed and failed results. If the provider accepted a refund and the response was lost, retain the reservation and pending receipt. Reconcile against the provider through the recorded reference using the mechanism supported by the approved adapter. Transition to confirmed only with authoritative evidence of the original refund and persist that result durably. Treat a stale or inconclusive lookup as pending. A definitive no-effect result is evaluated under the documented recovery policy and existing permissions. Never turn a timeout into an automatic second refund.

Resume from durable evidence

Start a continuation by rereading the latest session record and checking the actual working state. Recover any durable operation ID, reference, current state, and previous verification result. Match a completed test to its exact revision and environment before relying on it. If an earlier external effect may already exist, inspect authorized read-only status first. Preserve pending state and escalate the precise missing evidence if that check is unavailable. Persist the recovery result before further mutation.

Use this handoff record in the project repository or the designated local artifacts.

FieldExample
Scope and baselineProject Returns, PH-03, repository and branch, exact working revision, PRD-returns-r3, R-04 and R-02 evidence
Working stateEligible operation REF-0042 has its request reference and pending receipt
Verified checksNamed fixture, revision, environment, result and evidence path
Release gatesA-01 resolved at 45 days. Provider sandbox verification remains pending
Uncertain effectsTimeout after send. Preserve reservation and check original reference
Next actionReconcile REF-0042 using the authorized status fixture and verify exactly one observed effect
Owner and retry boundaryAdapter owner confirms authoritative lookup. No replay while outcome is unknown

Close the phase

Run the checks at the affected boundary and the appropriate surrounding regression suite. A skipped or unavailable check remains visible. For whole-requirements close-out, audit delivery against the current baseline across every required variant after implementation or completion work, then audit the tests. Reopen any implementation or evidence invalidated by that pass. Recheck assumptions, documented behavior and scope against the output. Review the diff for unrelated files, unintended generated outputs and sensitive values. Record the tested revision and any conflicts between the proposed diff and existing changes. Stage, commit, push, obtain credentials, publish or deploy only with the permissions governing that specific action. Hand off the evidence, affected criteria, and one safe next step, or record the terminal result when the unit is complete.

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 →