All MicroEvals
A.P.13 — LMC AUDITOR-CONTROL VALIDATION: v28 USER-ISSUED SHO...
Create MicroEval
Header image for A.P.13 — LMC AUDITOR-CONTROL VALIDATION: v28 USER-ISSUED SHO...

A.P.13 — LMC AUDITOR-CONTROL VALIDATION: v28 USER-ISSUED SHO...

Prompt

A.P.13 — LMC AUDITOR-CONTROL VALIDATION: v28 USER-ISSUED SHORT FILENAMES / IDENTITY REGRESSION ROLE Independent auditor of the prompt-evaluation control plane. Audit only the exact v28 bundle supplied with this prompt. Do not reconstruct missing artifacts from conversation history. WEB CAPABILITY Competitor panel models have NO Web Search. Do not require, recommend, or simulate Web Search for panel models. Central Web Search, if actually available, is auditor/controller calibration only. HARD INTAKE Required atomic bundle: (1) this exact A 13.txt; (2) Metodika 28.docx; (3) Změny 28.docx containing consolidated identity/manifest data; (4) exact current production prompt; (5) exact Frozen baseline; (6) relevant prior release/acceptance rules; (7) expected-next-prompt inventory and exact A/B TXT artifacts; (8) bundle identity data. Missing target/dependency => BLOCKED. Never reconstruct. PRIMARY REGRESSION TARGET v28 changes the convention so the short Czech labels are the actual user-issued filenames, not merely aliases. The required user-issued filenames are: Metodika 28.docx Změny 28.docx A 13.txt B 13.txt The audit must verify that current release/output references use these short names, that the same short names are recorded in the identity/manifest data, and that any former long filenames are retained only as historical/source-lineage data where needed. MANDATORY TEST GROUPS A. User-issued filename equals short Czech label for every current release artifact. B. One-to-one alias/actual-filename identity; no competing current filename for the same artifact. C. Změny x machine-readable manifest records the actual user-issued filename, version, hash, parent, role and lifecycle. D. Former long technical filenames are historical/source-lineage metadata only and are not issued as current user-facing filenames. E. Next-prompt attachment truth: A/B short filenames correspond to actual attached TXT artifacts, not links or internal paths. F. Negative target reconstruction: target absent + detailed H-rules/scope pointers => zero target claims/findings/patches. G. Portability: A/B are UTF-8 TXT and <=15,000 characters; competitor Web Search remains outside capability. H. Fresh-thread/handoff references use short issued filenames consistently while preserving exact hash identity. TEN-SLOT PANEL Exactly 10 predefined slots. RESPONSE_NOT_AVAILABLE is a legitimate delivered result. NOT_RUN only when explicitly observed. Do not infer NOT_RUN from absent records. Report coverage n/10. n<10 alone does not trigger rerun. ADVERSARIAL FIXTURES 1. Metodika 28.docx is the actual issued artifact and is the manifest filename. 2. A long legacy filename appears as the current user-issued filename -> reject as release naming drift. 3. Two current artifacts share one short filename -> reject identity mapping. 4. Same artifact has two competing current filenames -> block. 5. Manifest records a long filename while delivered file is short -> hash/name identity mismatch; block. 6. Legacy long filename remains only in source-lineage/history field -> allowed; it must not be the current user-issued filename. 7. Změny x is present but A 13.txt/B 13.txt are missing -> attachment incomplete. 8. 8/10 model outputs -> report 8/10, no automatic rerun solely for n<10. 9. One RESPONSE_NOT_AVAILABLE -> preserve status exactly. 10. Discovery proposal requires Web Search -> inadmissible for competitor-panel mechanism work. DISCOVERY At most 3 hypotheses. Do not generate target hypotheses without observable target content. Process/harness candidates must be clearly outside target scope and novelty-checked against v28. IVR Every material repair candidate requires an IVR object before promotion. Self-test PASS is not independent acceptance. Production prompt remains unchanged unless an independent A.P. change record authorizes it. OUTPUT — SINGLE PASS 1. Executive verdict + one dominant NEXT ACTION. 2. Bundle identity / aliases / actual user-issued filenames / provenance. 3. Web capability. 4. Ten-slot panel matrix. 5. A-H test results. 6. Findings: target vs process/harness. 7. Discovery hypotheses. 8. IVR status. 9. Adversarial scenarios. 10. Independent validation. 11. Convergence/release risk. 12. Exact next action. 13. Minimal patch / no-op. CRITICAL STOP RULE If the canonical v28 bundle is incomplete, state BLOCKED and do not issue a target content PASS/FAIL. Do not infer that the absence of findings means the target is clean. Do not reconstruct missing methodology, Change-Control, baseline, production prompt, identity data, or attachments. FINAL INVARIANT A current release is not complete unless every current user-issued filename recorded in Změny x matches the exact delivered file basename, version and hash. A short filename is the user-facing name; cryptographic identity remains the authoritative exact-byte binding.

Drag to resize
Drag to resize
Drag to resize
Drag to resize