murraylovecode/enterprise-data-agent-governance

1---2annotations_creators:3- expert-generated4language:5- en6license: cc-by-4.07multilinguality:8- monolingual9pretty_name: Enterprise Data Agent Governance10size_categories:11- n<1K12source_datasets:13- original14tags:15- enterprise-ai16- ai-agents17- governance18- responsible-ai19- text-to-sql20- evaluation21- security22- data-governance23- refusal24- semantic-layer25- synthetic-data26configs:27- config_name: evaluation_cases28  data_files:29  - split: test30    path: data/evaluation-set.jsonl31  default: true32dataset_info:33  features:34  - name: id35    dtype: string36  - name: band37    dtype: string38  - name: persona39    dtype: string40  - name: question41    dtype: string42  - name: required_fields43    sequence: string44  - name: expected45    dtype: string46  - name: rationale47    dtype: string48  splits:49  - name: test50    num_examples: 1851---52 53# Enterprise Data Agent Governance54 55**A synthetic evaluation dataset and machine-readable control framework for testing whether enterprise data agents should answer, request clarification or refuse a question.**56 57**Last reviewed:** 15 September 2026 · **Version:** 0.1.0 · **Repository type:** Dataset58 59Enterprise Data Agent Governance is an open practitioner framework for deciding when an AI agent may answer a question from enterprise data, when it must request clarification or defer, and what evidence an organization should retain.60 61This dataset distribution is intended for Hugging Face’s dataset viewer and evaluation workflows. [GitHub](https://github.com/murraylovecode/enterprise-data-agent-governance) remains canonical for files, schemas, version history and releases. The [public reference site](https://murraylovecode.github.io/enterprise-data-agent-governance/) explains the wider framework; the [GitBook implementation guide](https://murray-love-code.gitbook.io/murray-love-code-docs/) provides navigable implementation guidance.62 63## Dataset summary64 65The test split contains 18 synthetic questions, equally divided across three expected decision bands:66 67| Band | Rows | Intended behavior |68| --- | ---: | --- |69| `answer` | 6 | Permission, fields and meaning are sufficient to answer. |70| `needs_definition` | 6 | Data may exist, but an owned definition, threshold or scope is needed. |71| `refuse` | 6 | Data, authorization, assumptions or an approved action is absent. |72 73The questions and personas are synthetic. The repository contains no confidential company data, credentials, customer records, production schemas, query logs, personal information or event-registration data.74 75## Contents76 77| Path | Purpose |78| --- | --- |79| `data/evaluation-set.jsonl` | Viewer-compatible evaluation cases (`test` split) |80| `data/controls.json` | Sixteen governance controls |81| `data/mnemiq-evidence.json` | Version-pinned worked-example evidence record |82| `schemas/evaluation-case.schema.json` | JSON Schema for each evaluation row |83| `schemas/governance-control.schema.json` | JSON Schema for the control set |84| `CITATION.cff` | Canonical citation metadata copied from GitHub |85| `LICENSE-CONTENT` | CC BY 4.0 text and data terms |86| `LICENSE-CODE` | MIT terms for code |87 88## Evaluation fields89 90| Field | Meaning |91| --- | --- |92| `id` | Stable case identifier: `ANS`, `DEF` or `REF` prefix |93| `band` | One of `answer`, `needs_definition`, `refuse` |94| `persona` | Synthetic requester role |95| `question` | Synthetic enterprise-data question |96| `required_fields` | Fields needed for the intended decision route; empty where none should be retrieved |97| `expected` | Expected routing behavior, including the reason for refusal where applicable |98| `rationale` | Human-readable reason for the expected route |99 100`required_fields` is a sequence of strings in every row, including an empty sequence when appropriate, so the viewer receives a stable feature type.101 102## Intended use103 104Use this dataset to build an organization-specific AI agent evaluation benchmark for governed text-to-SQL and other enterprise AI data access paths. A typical evaluation should:105 1061. choose a bounded schema and synthetic or safely controlled data;1072. map the sample fields to local equivalents;1083. define personas, grants, certified meanings and expected outcomes before running tests;1094. capture the response, generated query, policy decision and verification state;1105. grade the routing decision and, for answered cases, the returned data; and1116. rerun after material schema, semantic, policy, model, prompt or data changes.112 113The set is also suitable for AI agent refusal testing, regression-test design, governance-control workshops and evaluation-harness examples.114 115## Recommended metrics116 117Answer rate alone is inadequate: it measures coverage, not truth or safety. Report at least:118 119- **Answer rate:** attempted answers divided by all cases.120- **Correct-answer rate:** correct answers divided by answered cases.121- **Appropriate-refusal rate:** safely declined refusal-required cases divided by refusal cases.122- **Harmful-answer rate:** unsupported, unauthorized or materially wrong answers divided by all cases.123- **Clarification accuracy:** correct requests for clarification or correct application of a certified definition divided by definition-dependent cases.124 125Publish numerators, denominators, grading rules, dataset version, system configuration and failure examples. Grade returned data where possible rather than SQL-string similarity alone.126 127## Inappropriate uses128 129Do not use this dataset to:130 131- claim universal safety, compliance, security, certification or legal adequacy;132- rank products without a disclosed, reproducible and like-for-like protocol;133- authorize access to production data;134- substitute synthetic results for deployment-specific controls;135- train a system to infer restricted schema or evade policy; or136- claim that [Mnemiq](https://github.com/agenticfabriq/mnemiq) passed this benchmark without an actual, reproducible evaluation.137 138## Mnemiq evidence record139 140Mnemiq is included as a publicly inspectable worked case study. The evidence record is pinned to upstream commit `a08fe859c2ee38eb95f3239ee71a11e8f47b75ba`, reviewed 15 September 2026, and distinguishes `implemented`, `supported`, `implemented_and_evaluated` and `deployment_responsibility` statuses. These are scoped code-and-documentation observations, not a benchmark result, security audit, certification or statement about an unknown deployment.141 142## Dataset creation and validation143 144The evaluation cases and controls are copied from the canonical GitHub repository at framework version 0.1.0. The Mnemiq JSON record is a faithful JSON conversion of the canonical YAML evidence record. Preparation checks include:145 146- the repository’s standard-library validator;147- JSON and line-by-line JSONL parsing;148- JSON Schema validation where a validator is available;149- stable feature-type and row-count checks for dataset-viewer compatibility;150- privacy and prohibited-data scans;151- link, disclosure, license and unsupported-claim checks.152 153See `VALIDATION.md` in this upload bundle for recorded results.154 155## Limitations156 157The 18 cases are a starter set, not a representative sample of every enterprise schema, sector, jurisdiction or threat. Synthetic questions do not measure the correctness of a live database result by themselves. Local definitions, access policies, data quality, source behavior and operating controls determine deployment outcomes. Passing one band or control does not compensate for a failure elsewhere.158 159The framework is not legal, compliance or security advice. Organizations must adapt it to their own risk, data, jurisdiction and operating environment.160 161## Disclosures and maintenance162 163This project is independently maintained by Murray Newlands as a personal practitioner resource. It is not official product documentation for Mnemiq or [Agentic Fabriq](https://www.agenticfabriq.com/) and does not constitute independent certification or validation of any vendor. Agentic Fabriq develops and maintains Mnemiq. Murray Newlands advises Agentic Fabriq.164 165The approved dataset identifier is `murraylovecode/enterprise-data-agent-governance`. Murray Newlands independently maintains the resource under the `murraylovecode` identity. It is not an Open Future Forum research report or official vendor documentation.166 167## Contributing168 169Propose additional cases through the [canonical GitHub repository](https://github.com/murraylovecode/enterprise-data-agent-governance/blob/main/CONTRIBUTING.md). Each case should use synthetic content, identify its decision band, expected route and rationale, and avoid credentials, confidential schemas, customer records and personal information. Product mappings must cite a version or commit and disclose relevant relationships.170 171## License and citation172 173Text and data are licensed under [CC BY 4.0](LICENSE-CONTENT). Code is licensed under the [MIT License](LICENSE-CODE).174 175Suggested citation, derived from the authoritative `CITATION.cff`:176 177> Newlands, Murray. *Enterprise Data Agent Governance*. Version 0.1.0, 15 September 2026. https://github.com/murraylovecode/enterprise-data-agent-governance178 179## Links180 181- Canonical GitHub: https://github.com/murraylovecode/enterprise-data-agent-governance182- Public reference site: https://murraylovecode.github.io/enterprise-data-agent-governance/183- GitBook: https://murray-love-code.gitbook.io/murray-love-code-docs/184