OMDQS

Prompt

Rewrite and restructure the following prompt to be an optimized prompt template that contains an {{input}} placeholder and provide any variants of the template and a broad spectrum one to cover all use cases - including but not limited to a single question with multiple provided options that the optimal option needs to be selected (one quality space, one decision), a single question with sometimes some options provided and the output should take into account both provided options and other options that were not presented or that combine different options (one quality space, one decision), a single situation with multiple questions or queries where the questions are discrete in the sense that the optimal outcome is the combination of selecting the optimal answer to all the questions collectively (one quality space, one decision, many given and many not given options per query where the optimal outcome is the perfect mix that synergizes), a single quality space with multiple discrete queries, where the optimal answer to each query must use all available and other options that could be better than the provided options for each one, multiple different but related quality spaces each with one or more queries each of which provide multiple stated options and multiple not enumerated but needs to be discovered through thinking beyond provided answers and trying to combine answers, and any other template use case that is along these lines - you must provide multiple templates, a single template that handles all these cases and 3 meta-templates that use recursive meta prompting to come up with the prompt below for that situation, use a quality space to create a recursive meta prompt that is optimal and then execute it and the third uses recursive meta prompting in a manner that uses the method provided in the quality space prompt below in order to develop the optimal prompt and then to execute that prompt to arrive at the desired solution(s). If you are unsure about any ambiguities, you must just provide enough prompt templates that can be customized enough that collectively every possible use case you can think of along these lines is achieved: *** Using the information or problem space provided below, along with the other information you have available to you using a variety of internal and external sources, give all items a weighted orthogonal higher dimensional quality space co-ordinate with the space scaling proportionally to the optimal decision and each near-orthogonal factor must be a dimension. Then calculate the distance between each option below and in the attached document and the optimal point to give me the score each option earned. Example Input 1: ### 1. Executive Summary & Problem Restatement **The Core Question:** When multiple documents across the Paste Verdict Engine (PVE) project—and across the boundary between the workstation estate and PVE—contain overlapping statements, configurations, or facts (e.g., operator identity/rules, skill counts, engine version numbers, function contracts, model availability), **how does PVE know, at any given moment, the single authoritative home for every fact, and what mechanical process enforces and audits that contract rather than merely writing it down?** This problem spans two distinct operational boundaries: 1. **PVE-Internal Core (31 Managed Files):** Where facts like engine rules, function contracts, and per-file version numbers exist. (e.g., `STATUS.md` already owns the per-artefact version table). 2. **Cross-Estate Boundary (The Imported Estate Files):** Where facts originating on the workstation (`D:\Context\fact-ownership.json`, `CONTEXT_SSOT.md`, `CLAUDE.md`) are imported into PVE as read-only dependencies, spanning a cloud Linux VM and a local Windows machine. **What IS Being Asked:** * What architectural mechanism governs fact ownership across PVE's files? * Does PVE instantiate a dedicated registry file, embed ownership in frontmatter, rely on a structural convention (single-home policy without a registry), or externalize tracking to automated build tooling? * How does PVE manage the boundary between workstation estate facts and cloud project files across its dual execution profiles (standard cloud web vs. Perplexity Computer with device bridge)? * What active consumer mechanically enforces this contract so the registry does not rot into dead documentation? * How does the system detect out-of-band single-file web uploads and prevent container-scoped discrepancies? **What is NOT Being Asked:** * We are **not** questioning whether singular fact ownership is the right principle (settled canon under R6, C46, and Constitutional Principles 82/83: every duplicated fact has exactly one owner). * We are **not** guessing or asserting fact counts; empirical measurements have replaced speculative estimates. * We are **not** proposing to alter the workstation's central `D:\Context\fact-ownership.json`. * We are **not** deciding to merge, rewrite, or delete any specific project files yet. --- ### 3. Actors, Stakeholders, and Components Involved | Actor / Component | Operational Role in Fact Ownership | Grounding Canon | | :--- | :--- | :--- | | **Human Operator (Hersh)** | Sole approver; resolves canon conflicts; mandates the **Burden Ban** (no unnecessary AI ceremony, zero ritualistic maintenance, plain-English visibility). | `CONFIG_PLAYBOOK.md` | | **Standard PVE Runtime Agent** | Executes in cloud Linux VM; must resolve authority using only uploaded files without workstation reach. | `ADAPTERS.md` | | **Perplexity Computer Runtime Agent** | Operates on-demand device bridge (`pc` commands to `Hersh-PC`); can audit live files; not ambient. | `ADAPTERS.md`, Q10 Ruling | | **Local Packaging / Build Agents** | Workstation agents (CLI, CC, Antigravity) that execute the pre-upload build gate before a zip is generated. | New Build Gate | | **`STATUS.md`** | Already the authoritative canonical owner of the per-artefact version table and content counts. | `STATUS.md`, `READ_ME_FIRST.txt` | | **`CONFIG_PLAYBOOK.md`** | Existing deterministic enforcement layer: single-home rule (Check 7) and manifest completeness (Check 9). | `CONFIG_PLAYBOOK.md` | | **`CHALLENGE_TIERS.md`** | Governs how a detected conflict between a paste and a file, or between two files, gets flagged rather than silently resolved. | `CHALLENGE_TIERS.md` | | **`ESTATE_FILES.md` & Estate Items** | The boundary population itself: read-only dependencies PVE must never edit or regenerate. | `ESTATE_FILES.md` | | **`REJECTED_IDEAS.md`** | Prior rulings against a live document connector for estate files (R5), instruction field cryptographic fingerprints (R9), and standing drift-history logs (R10). | `REJECTED_IDEAS.md` | | **Workstation Estate Authority** | External authority (`D:\Context\fact-ownership.json`) that PVE inter-operates with but never absorbs. | `fact-ownership.json` | --- ### 4. The Seven Dimensions of the Problem Space 1. **Locus of Enforcement:** Where does the check run? (Pre-upload build script, in-runtime agent reasoning, cloud manifest check, or human review). 2. **Fact Categorization & Operational Invariants:** How are facts classified by their operational behavior (computed vs. per-file declared vs. boundary vs. prose)? 3. **Detection & Verification Method:** How is drift discovered? (Explicit enumeration, raw SHA-256 content hashing, semantic fact fingerprinting, or similarity search [rejected]). 4. **Temporal & Staleness Model:** How is decay tracked? (Per-fact days threshold, event-triggered version counter, ambient bridge sync [rejected], or advisory-by-default). 5. **Boundary-Crossing Strategy:** How are estate facts handled in cloud storage? (Pull model, push model, hash-anchored mirror, or anchor-summary pointer). 6. **Container-Scoped Semantics:** How are statements scoped so they remain true across container boundaries (`delivery-zip` vs. `project-container`)? 7. **Failure Posture & Escalation:** What happens when drift or staleness is detected? (Hard build halt, soft warning banner, `CHALLENGE_TIERS.md` basis-request, or operator override). --- ### 6. Summary of Architectural Directions * **Direction 1: Four-Tier Model with Sequenced Build Gate & Cloud Manifest (Recommended)** * Sequence: Build the packaging script first; co-deliver `FACT_REGISTRY.json` covering the 7 boundary files. * Invariants: Tier 1a derived live; Tier 1b checked per-file against `STATUS.md`; Tier 2 checked via raw SHA-256 with per-fact thresholds; Tier 3 governed by anchor-summary pointers. * Runtime protection: Shipped `FILE_MANIFEST.sha256` detects ad-hoc single-file web uploads in cloud sessions. * Scope: Statements carry explicit container scopes (`delivery-zip` vs. `project-container`). * **Direction 2: Pure Pointer Convention with Anchor Summaries (Zero Registry Files)** * Ban fact restatements entirely across all PVE files; every secondary reference must be an explicit citation accompanied by a one-line summary. * Zero registry files created, but leaves cross-boundary estate staleness entirely to manual tracking. * **Direction 3: Externalized Workstation-Only Registry** * PVE holds zero registry artifacts in the cloud. * PVE files are tracked solely inside [`D:\Context\fact-ownership.json`](file:///D:/Context/fact-ownership.json) on `Hersh-PC`. * Pre-upload scripts validate files locally; cloud runtime has zero visibility into ownership metadata. * **Direction 4: Live Bridge-Dependent Auditing** * PVE relies actively on Perplexity Computer `pc` commands to query `Hersh-PC` during Config turns. * Incompatible with standard mobile/web sessions where the bridge is unavailable. * **Direction 5: Push-Based Estate Delivery with Advisory Tagging** * Invert the pull model: the workstation estate push-scripts fresh files whenever an estate rule changes. * If un-updated, PVE marks all estate-derived statements as advisory/historical by default, removing the expectation of runtime freshness. --- ### 7. Clarifying Questions for Operator Ruling (Decision #7) To move from this mapped problem space to a final, binding custom decision, the following specific choices remain for the operator: 1. **Adoption of the Refined Partition Model:** * Do you approve partitioning facts into: * *Tier 1a (Computed live from content)* * *Tier 1b (Per-file version check against `STATUS.md` table)* * *Tier 2 (The 7 estate boundary files with raw SHA-256 and per-fact staleness thresholds)* * *Tier 3 (Native prose with anchor-summary pointers)*? 2. **Implementation Sequencing:** * Do you agree that developing the **local packaging/build script** is a prerequisite to (or co-delivered with) creating `FACT_REGISTRY.json`, so the registry is never born in an un-enforced passive state? 3. **Single-File Upload Bypass & Manifest Detection:** * Do you approve packaging a `FILE_MANIFEST.sha256` in the project export so the cloud runtime can detect ad-hoc single-file browser uploads without needing the device bridge? 4. **Failure Posture on Cloud Manifest Mismatch:** * When the cloud runtime detects that an ad-hoc single-file upload has occurred (hash mismatch against `FILE_MANIFEST.sha256`), should it: * *(Option A)* Emit a single-sentence warning banner to Hersh and proceed? * *(Option B)* Route through the `CHALLENGE_TIERS.md` Tier 2 basis-request pattern (non-blocking, logged assumption)? * *(Option C)* Halt the turn and require re-exporting the full package? 5. **Re-harvest Trigger & `#config` Keyword:** * Do you confirm that updating an estate snapshot or updating `owner_sha256_at_harvest` is classified as a configuration change that strictly requires the `#config` keyword (consistent with R1)? 6. **Sanctioned Override Alignment:** * Do you approve that the `--override-stale` flag follows the Decision #10 override pattern (operator retains final call, warning emitted, confirmation phrase required for high-stakes releases)? 7. **Perplexity Computer Device Bridge Role:** * Should the device bridge remain strictly an **opportunistic accelerator** (used during Computer turns to verify harvest hashes against `Hersh-PC`), while keeping PVE’s core files completely self-contained for standard mobile/web chat?