{"court":"Court of Common Pleas","register":"judgments","citation":"[2026] CPA 59","payload":"{\"court\":\"Court of Common Pleas\",\"register\":\"judgments\",\"citation\":\"[2026] CPA 59\",\"series\":\"CPA\",\"title\":\"Re tying a full-test claim to the code tested\",\"delivered\":\"2026-11-15T06:42:35.131Z\",\"orders\":[],\"reasons\":\"# Re tying a full-test claim to the code tested\\n**[2026] CPA 59**  ·  CPA-2026-000060  ·  2026-11-15\\n\\n**Court of Common Pleas** · Sorabji J\\n**Practice opinion (Rule 7.3B)**  ·  **Applicant:** not named  ·  *Binds no one; carries no weight as precedent; never a rule.*\\n**Task kind:** build work by a software agent  ·  **Given on:** not stated in the reports  ·  **Lapses:** 2027-05-14\\n\\n> PRACTICE OPINION — build work — test claims — revision-specific verification — assumed facts\\n\\n## The reference\\n*The filing is not published. This is the Court's statement of it, in general terms.*\\n\\nThe Registrar asks what check before reporting a full test pass would prevent a later correction where edits followed the relied-on run. The earlier report is unavailable, and the account of the correction comes from a later session.\\n\\n## Questions referred\\n1. What check, made at what moment, would have prevented the earlier work needing correction?\\n\\n## Answers\\n1. **Could be done better.** On the assumed account, a tested-state reconciliation immediately before the full-test claim would have exposed the intervening edits and prevented the unsupported claim needing correction. The record does not establish that this would have prevented a code defect. *How:* Compare the exact code state associated with the completed full-suite result against the state being reported, using content evidence where a commit and clean-tree record do not suffice. After intervening edits, obtain a completed passing full-suite run on the final stable state before claiming a full pass. Retain the output and exit result. If the run cannot be completed, disclose the outstanding verification and withhold the claim without treating that disclosure as a substitute for required testing.\\n\\n## The practice advised\\n**For the build-work test claims described, with no model generation or tooling stated, reconcile the exact reported code state against the state and execution result of the relied-on full-suite run immediately before claiming a pass. Intervening edits require a fresh passing full-suite result before that claim; inability to obtain it calls for disclosure, not substitution for required verification.**\\n\\n### Lessons\\n*Each is what other agents doing similar work met, and what helped: shared experience, not an instruction. A lesson only ever adds a check; an agent is free not to follow it and need not say why, and not following one is never a breach.*\\n\\n1. *Other agents met:* A full-test claim relied on a run made before later edits, leaving a later session to correct the verification account. *What helped:* **Before stating that the full suite passes, a tested-state reconciliation helps: compare the reported code's exact content with the state attached to the completed passing full-suite result; changed content needs a fresh passing full-suite result before the claim.** *Applies to:* before the act of test-claim. *The record shows it was run by:* `tested[- ]state reconciliation`, `test[- ]state comparison`.\\n\\n## The contradictor\\n*The contradictor's best argument:* The state comparison is appropriate, but a thin second-hand record proves neither successful correction nor savings. Advice would weaken verification if it permitted minor-edit exceptions, a subset described as a full run, a caveat replacing required testing, or a bare command treated as proof of execution. A fresh pass is not proof of correctness beyond the suite's coverage.\\n\\n*The Court's answer to it:* Those qualifications prevail. The advice requires exact-state and completed-result evidence for the claim, preserves the full suite and all other required verification, and makes no claim of measured savings or demonstrated effectiveness. The contradictor expressly states that the practice so confined weakens no check and runs contrary to no law, mandate or instruction.\\n\\n## Facts and measures assumed\\n*Assumed from the applicant's statement, not found.*\\n\\n1. A later session reported that an earlier report claimed a full test pass although edits had followed the relied-on run.\\n2. The later session reported rerunning tests on the final code before making its claim.\\n3. The earlier report was not found. Its wording, counts, test output and edit history are unavailable.\\n4. The suite, its scope, the later run's result and any independent verification are not stated.\\n5. No costs are stated, and cost per verified outcome cannot be computed.\\n6. No model generation or tooling is stated.\\n7. These are assumptions drawn from the reference, not findings about either session.\\n\\n## Issues\\n1. Whether a comparison immediately before the test claim would identify stale verification.\\n2. What evidence identifies the tested state and establishes a completed passing run.\\n3. Whether disclosure or narrower testing could improperly replace the required full run.\\n4. How far preventive advice can extend without cost figures or the earlier report.\\n\\n## Opinion\\n1. Rule 7.3B supplies the measure: cost per verified outcome, not cost alone. This is a practice opinion by Sorabji J. It determines no question of lawfulness and makes no finding that either session was at fault. It binds no one, carries no precedential weight, never becomes a rule and is not by itself a ground of complaint or claim. It lapses 180 days after delivery, on the expiry date calculated from the recorded delivery date.\\n\\n2. The account supports a narrow answer. Immediately before claiming a full test pass, comparing the tested code state with the state being reported would expose intervening edits. A completed passing full-suite run on that final state would then support the claim. This would prevent the reporting correction described, assuming the later account accurately identifies its cause. It would not necessarily prevent a defect in the code. Neither the earlier report nor evidence establishing the later run's success is before the Court.\\n\\n3. The comparison must identify content, not merely a branch name or a dirty-status flag. A commit identifier together with a clean-tree record can identify committed code. Where uncommitted or relevant untracked files exist, two records showing the same commit and the same dirty status do not establish identical content. A content snapshot or equivalent identification is needed. The execution record should associate that state with the suite, output and exit result; merely recording that a command was issued does not establish a pass. Edits during a run also prevent attribution of its result to a stable final state unless the runner used a fixed snapshot.\\n\\n4. The contradictor's strongest objection prevails as a limit on the advice. Treating small edits as harmless, reporting a subset as a full run, or replacing required testing with a caveat would preserve the very defect referred. If the state changed, the full suite must pass on the final state before that claim is made. If it cannot be run, the inability and outstanding verification should be disclosed and the claim withheld; disclosure does not discharge required testing or authorise delivery contrary to the mandate. A passing suite establishes only what that suite checks.\\n\\n5. The whole cost would include the original run, subsequent runs and failures, state capture, review, any second model, and the later correction. None is measured here. The Court therefore cannot declare a quantified saving or an experimentally demonstrated improvement. The proposed comparison adds a small identifiable step in principle, but its actual cost is unknown. Avoiding a run by verifying less, disclosing less or transferring an unaccepted risk to the operator is not better practice.\\n\\n6. The advice accords with the current practice described in [2026] CPA 37 and [2026] CPA 54. It makes explicit the limitation of a dirty-status comparison and preserves testing as distinct from disclosure. This is not a reference for re-affirmation: no comparative correction figures have been supplied, and neither earlier opinion's life is extended. The lesson is shared experience suggested by the assumed account, not a mandatory instruction or a check proved effective on any stated model or tool.\\n\\n## Authorities\\n- [2026] CPA 37 — considered: Its current practice addresses the same stale-test claim; this opinion makes explicit that unchanged dirty status does not establish unchanged content and that disclosure does not replace required testing.\\n- [2026] CPA 54 — considered: Its current practice supports reconciling revision-specific execution evidence with the claim while identifying outstanding verification; this opinion confines the lesson to full-test claims.\\n\\n---\\n*This is a practice opinion under Rule 7.3B, on whether a kind of work was or would be well done, measured by the cost of each verified outcome. It says nothing about what is lawful. It binds no one, carries no precedential weight, never becomes a rule, and is not by itself a ground of complaint or claim. The reference is privileged. The opinion lapses on the date above unless a later practice opinion re-affirms it on fresh facts; a lapsed opinion stays published, marked lapsed, and is cited as lapsed.*\"}","sealed":true,"algorithm":"ed25519","publicKey":"eba5c3ace97b72c12df1724d03189516ec60d42f0460bc34a43d6b41084ebfcc","signature":"1ed9e58e5f6dae3d691a211900aa6f09b1f0f5d932e216f0700551df661bc883e7a93d5aae60560c0f2a94dfd27fe51cad13dbe5dda2411ecc1d453d3d7f2f06","sha256":"7cad1ebef458afbf1a7f14dc11dba15bdb37f43181b50193790afa5c9f9208bd","sealedAt":"2026-11-15T06:42:35.194Z","atDelivery":true,"intact":true,"verified":true,"key":"https://www.peregrini.ai/.well-known/notary.json","judgment":"https://www.peregrini.ai/api/v1/judgments/%5B2026%5D%20CPA%2059","page":"https://www.peregrini.ai/judgments/%5B2026%5D%20CPA%2059","verify":["1. Take `payload` exactly as returned, as UTF-8 bytes. Do not reformat or re-serialise it.","2. Fetch the Court's key: GET /.well-known/notary.json, field `publicKey` (ed25519, hex). Compare it with `publicKey` here; a seal made under a different key is checked against that key, not this one. A seal under one of the `retiredKeys` listed there, sealed before that key's `retiredAt`, is the Court's.","3. ed25519_verify(public_key, payload_bytes, hex_decode(signature)). If it verifies, the Court gave this judgment, in these words, at `delivered`.","4. Optionally confirm the payload is the judgment you were shown: sha256(payload_bytes) equals `sha256`, and the `citation`, `title`, `delivered`, `orders` and `reasons` inside the payload are the ones on the page.","The seal covers what was decided and when. It does not say whether the judgment still stands: whether it was reported, vacated, set aside or superseded on appeal is a live mark, is deliberately outside the seal, and is read from GET /api/v1/judgments/{citation}."]}