Skip to content

Verification Report Template

The verification report is Tester’s concise, resumable assessment for one cycle. Use the location and provenance defined by Workflow Artifact Provenance in the protocol. Tests remain in the repository’s established test locations. Both modes use this format; do not create a separate preferences or status file.

Create the report with node .standards/bin/artifact.mjs init VERIFICATION, which writes this provenance block and the header below:

<!-- STANDARDS
Artifact: VERIFICATION
Cycle: <Active Work.Id>
-->

Cycle: <Active Work.Id> Mode: VERIFY | REVERIFY Status: IN_PROGRESS | BLOCKED | COMPLETE User Style: NONE | <identifier> Assessment Purpose: FULL Assessment Target: NONE

  • Scope, architecture, context, and development-plan paths with relevant revisions/content identities; current acceptance set.
  • Active-change source and test boundaries, including committed and dirty-tree content, relevant dependencies/configuration, and execution environment.
  • Reason for this assessment or reconciliation; session-freshness visibility limitation when applicable. Do not claim a machine-certified fresh session.
  • Current assessment purpose and target, the relevant increment definition or corrective assignment, and the exact implemented outcome being assessed.
AC / technical criterion Tests or checks Disposition and evidence
AC-NNN / design section Scenario/check reference Result or gap reference

Use test paths and scenario names, commands, or inspections. Record verified with result evidence, blocked/uncovered with the gap, or pending a later role with its owner and required evidence.

For partial assignments, use AWAITING_IMPLEMENTATION and distinguish partial AC evidence under the protocol’s Acceptance Traceability.

Account for all current IDs and relevant technical criteria without rewriting their meaning. Separate criteria sharing an ID when their evidence differs. Retired references may remain in a concise superseded-evidence note, never as current coverage. Presence of a test is not a passing result.

Include this section once an increment has been assigned; retain it across cadence switches and later full verification.

Increment Outcome and acceptance scope Assessed inputs / evidence Disposition
1 Defined outcome / AC-NNN Content / result pointers VERIFIED, BLOCKED, or SUPERSEDED

Reference the Developer-owned increment definition and the evidence required by Checkpoint Handoffs. Preserve useful assessment history when outcomes are revisited or inputs change. Keep current acceptance conclusions in Acceptance Evidence; this table records assignments and their results.

Source file Reused coverage Active-change allocations User additions
Source Reused scenarios Allocations Additions

Record any user additions with their targeted area/limit and user direction. Apply the counting rules in SKILL.md; record cross-file assignments and parameter/property counts when relevant. Preserve allocations across sessions, modes, increments, and cadence switches, including invalidated/replaced or removed scenarios. Do not list every unrelated existing test. When blocked by the ceiling, name the unmet criterion, proposed additional scenarios, and smallest requested increase.

  • Exact command and working directory (or precise inspection performed).
  • Assessed revision/content and relevant environment/configuration; when run.
  • Actual result, exit status if available, counts and log/artifact pointers when useful. Separate passes, failures, skips, and unrun/unavailable checks.
  • Coverage/limitations: what the result establishes and what it does not.
  • Still-valid prior Tester result reused and reason, or superseded result and why new evidence is required. Developer self-checks are not formal evidence.

Keep only enough detail to audit and resume; link useful logs rather than copying terminal transcripts. Never record credentials or sensitive values.

  • Evidence/reproducer, affected AC or technical criterion, owning role, and required correction for each unresolved defect.
  • Required unrun checks or uncovered obligations and the cause of each gap.
  • Explicit later-role dependency, its AC, owner, and evidence still required.
  • Blocking user question, if any, matching Active Work.BlockedOn.

Record the next concrete action on interruption, or the completion conclusion and evidence locations for the receiving role. State which acceptance evidence remains pending downstream. Supply enough context for the protocol’s independent Reviewer session and identify the requested review kind. Workflow state and recovery routing remain canonical in STATE.md, not in this report.

Store suspended assignments, scoped correction evidence, and remaining work here under the protocol’s Implementation and Verification Recovery Gates.


Keep Assessed Inputs, Acceptance Evidence, Scenario Budget, Execution Evidence, and Resume or Handoff. Omit Open Findings and Dependencies only when empty. Use explicit NONE or not-yet-run explanations rather than empty required tables.

Set User Style to the user’s explicit selection under the protocol’s User Styles, or NONE.

Set Assessment Purpose independently of Mode:

  • FULL: reconcile the entire current implementation contract; target NONE.
  • INCREMENT: assess the assigned testable outcome and affected completed behavior; target Increment N, matching the increment assigned at handoff. After the return, retain that target as history until a new assessment starts, even if Developer changes Current Increment or cadence.
  • CORRECTION: verify a specific owned correction or affected recovery rerun; target a concise description of that assignment. Record its frame association and interrupted assignment in Resume or Handoff when recovery is active.

Assessment Purpose and Assessment Target are required. Initialize new reports with FULL and NONE, respectively. Missing, invalid, repeated, or contradictory values are inconsistencies. An increment or correction requires its explicit target and scoped evidence before its gate can pass. Use VERIFY or REVERIFY according to the existing input-change rules; purpose and cadence do not reset those obligations.

IN_PROGRESS means verification is underway; BLOCKED means required work cannot currently complete. Neither declares full completion, although an IN_PROGRESS report may record a passed increment or correction gate. COMPLETE requires purpose FULL and the full shared Tester gate, possibly with explicitly permitted later dependencies, under Full Verification Boundary. Reopen a completed report when reconciliation is needed. Keep current evidence coherent, preserving useful result and allocation history.