Acceptance Criteria for Per-session Traceability From Instruction to Human Decision: What Must Be Proven

The direct answer

Write a test for “every assigned instruction has a handling identity” before execution and keep “identifiers regenerated between tools” as a release-blocking counterexample.

For per-session traceability from instruction to human decision, the relevant audience is teams that cannot reliably connect agent assignments, produced artifacts, QA verdicts, and issue-resolution decisions. The decision should cover stable session identity, instruction assignment, agent identity, tool and artifact references, QA verdicts, exceptions, human decisions, and closeout coverage. The supplied boundary starts with the current agent setup and representative session logs and ends with a per-session run directory and traceability coverage check, presented in reviewable form.

Trace completeness supports review; it does not prove correctness, approval, compliance, or the truth of an artifact's claims.

Turn each requirement into a proof obligation

The expected deliverable is a per-session run directory and traceability coverage check.

Use the current agent setup and representative session logs as the controlled starting material.

The QA reviewer adjudicates every pass, hold, or fail verdict against these registered statements.

Cover more than the happy path

The normal flow should establish whether “every assigned instruction has a handling identity” holds. An alternate flow should vary a permitted input while testing whether “artifacts and tool receipts are addressable” holds. The failure flow should use a fixture demonstrating “QA verdicts recorded without proving output” and verify containment.

Add a recovery flow for “human decisions captured without rationale”.

Judge evidence quality and freshness

For per-session traceability from instruction to human decision, a result from another environment cannot prove that “QA verdicts include evidence” holds in the buyer's environment.

Define pass, hold, and fail before execution

Keep sign-off independent

The implementer may produce artifacts, but the QA reviewer should judge whether “secrets and unnecessary personal data are excluded” holds against criteria written before the result was seen.

Record the business decision of the session owner, the technical evidence reviewed by the tool operator, the acceptance verdict recorded by the QA reviewer, and residual risk accepted by the human decision owner.

A screenshot or self-score cannot prove that “secrets and unnecessary personal data are excluded” holds under the failure condition “sensitive values copied into the audit record”.

Reopen criteria when the system changes

Changes to stable session identity, instruction assignment, agent identity, tool and artifact references, QA verdicts, exceptions, human decisions, and closeout coverage can invalidate a test even when the requirement text stays the same.

How the sources bound the acceptance criteria decision

For per-session traceability from instruction to human decision, the live catalog limits the offer to two elements. The supplied boundary is the current agent setup and representative session logs. The catalog names the deliverable as a per-session run directory and traceability coverage check. It cannot establish whether “every assigned instruction has a handling identity” holds in the buyer's environment.

Connect those narrow roles to a local fixture for “artifacts stored without the instruction that produced them” rather than treating citation status as a pass.

For per-session traceability from instruction to human decision, limit the conclusion to the documented workflow and let the agent supervisor retain the current source-to-claim map. Reopen the source judgment if the failure case “identifiers regenerated between tools” changes the tested conditions.

Product-specific acceptance criteria review drills

These drills connect per-session traceability from instruction to human decision to concrete inputs, failures, acceptance statements, and owners. For per-session traceability from instruction to human decision, the drills map each criterion to a reviewable verdict.

Acceptance for per-session traceability from instruction to human decision is judged against a boundary record covering the current agent setup and representative session logs, never live protected material. The session owner requires synthetic, non-secret cases; messages, writes, state changes, and all other external effects stay inside the fixture throughout and after each case.

Requirement trace

Reproduce a safe case involving “identifiers regenerated between tools” as the entry condition for the requirement trace review. The session owner preserves the last state that the workflow can prove.

Run the case within the documented boundary covering the current agent setup and representative session logs while the agent supervisor checks whether “QA verdicts include evidence” holds. The observation must come from outside the candidate's self-report.

The QA reviewer advances the record only when it can demonstrate “QA verdicts include evidence”. If evidence conflicts, the QA reviewer records fail and preserves the prior state. In the requirement trace review, evidence for “QA verdicts include evidence” maps support to pass, contradiction to fail, and unresolved to hold.

The QA reviewer reopens the drill if the criterion “QA verdicts include evidence” is judged with a different fixture, policy, or operating state.

Normal-flow result

Describe the normal-flow result review through a case involving “artifacts stored without the instruction that produced them”. The agent supervisor captures the known state and the first unanswered workflow question.

Compare the candidate result with a frozen scope record covering the current agent setup and representative session logs for “secrets and unnecessary personal data are excluded”. Preserve both sides of the comparison.

If the case establishes “secrets and unnecessary personal data are excluded”, the QA reviewer authorizes the next limited action. Unresolved evidence keeps a per-session run directory and traceability coverage check on hold; contradictory evidence makes the QA reviewer record fail. In the normal-flow result review, evidence for “secrets and unnecessary personal data are excluded” maps support to pass, contradiction to fail, and unresolved to hold.

Reopen this result after a change to the input, the authority of the agent supervisor, or the workflow condition represented by “artifacts stored without the instruction that produced them”.

Alternate-flow result

Begin with the adverse condition “QA verdicts recorded without proving output”. During the acceptance criteria review, the tool operator locates its first observable effect inside stable session identity, instruction assignment, agent identity, tool and artifact references, QA verdicts, exceptions, human decisions, and closeout coverage.

The proof package identifies the input boundary as the current agent setup and representative session logs and includes a direct check that “artifacts and tool receipts are addressable” holds. Assumptions stay separate from observed artifacts.

Let the QA reviewer decide whether the criterion “artifacts and tool receipts are addressable” passed under the recorded conditions. That verdict controls only this review slice. In the alternate-flow result review, evidence for “artifacts and tool receipts are addressable” maps support to pass, contradiction to fail, and unresolved to hold.

Recheck the drill when the operating path no longer matches stable session identity, instruction assignment, agent identity, tool and artifact references, QA verdicts, exceptions, human decisions, and closeout coverage or when the rollback evidence expires.

Failure-flow result

Frame the failure-flow result review around “human decisions captured without rationale”. Before testing a response, the human decision owner captures the input, decision boundary, and residual state.

Retain a boundary record covering the current agent setup and representative session logs, the observed output, and the test for “open issues resolve to a named decision”. This makes the decision reproducible.

The QA reviewer resolves the drill with one finding about “open issues resolve to a named decision”. For per-session traceability from instruction to human decision, the deliverable decision in the failure-flow result review advances only when that finding is supported. In the failure-flow result review, evidence for “open issues resolve to a named decision” maps support to pass, contradiction to fail, and unresolved to hold.

Do not reuse the disposition when the failure case “human decisions captured without rationale” occurs under conditions outside the recorded input and authority boundary.

Independent verdict

Attach a fixture for “sensitive values copied into the audit record” to the independent verdict review decision record. The human decision owner marks the exact point where human review becomes necessary.

Review the scope record covering the current agent setup and representative session logs under its recorded authority and evaluate whether “every assigned instruction has a handling identity” holds. The session owner owns the evidence gap.

The QA reviewer moves forward only after the record supports the finding “every assigned instruction has a handling identity”. Conflicting evidence makes the QA reviewer record fail and preserve the prior state. In the independent verdict review, evidence for “every assigned instruction has a handling identity” maps support to pass, contradiction to fail, and unresolved to hold.

Return to the independent verdict review after a dependency change alters the path from “sensitive values copied into the audit record” to the reviewed end state.

Evidence expiry

Open an evidence expiry review record for the failure case “identifiers regenerated between tools”. The session owner maps the trigger to one reviewable transition in stable session identity, instruction assignment, agent identity, tool and artifact references, QA verdicts, exceptions, human decisions, and closeout coverage.

The agent supervisor receives a boundary record covering the current agent setup and representative session logs with an explicit request to verify whether “QA verdicts include evidence” holds. Input identity and judgment stay in the same receipt.

The QA reviewer treats completion as insufficient unless the record resolves “QA verdicts include evidence”. Merely producing a per-session run directory and traceability coverage check does not settle the drill. In the evidence expiry review, evidence for “QA verdicts include evidence” maps support to pass, contradiction to fail, and unresolved to hold.

Revisit the evidence expiry review after an input, owner, or consequence change invalidates the proof that “QA verdicts include evidence” holds.

Frequently asked question

What acceptance criteria should I use for Agent Audit Trail?

Require observable evidence that every assigned instruction has a handling identity and include “identifiers regenerated between tools” as a negative case. The QA reviewer should record pass, hold, or fail before expansion.

A product bridge, with a boundary

The Agent Audit Trail is the relevant sincLLM offer for this narrow problem. The frozen live catalog describes its required boundary as the current agent setup and representative session logs and its deliverable as a per-session run directory and traceability coverage check. Delivery under the catalog scope cannot by itself prove buyer fit, legal compliance, system safety, technical adequacy, or a business outcome.