Skip to content

cd-ux/hse-agent-public

v1.0.19CC-BY-4.0

Domain-adaptable Human-Systems Engineering guidance for product, research, workflows, requirements and validation.

GitHubOpen Website

hse-agent — Public Edition

By Christopher Durgin · 1.0.19 · Public edition

Reusable Human-Systems Engineering guidance for research, product strategy, workflows, decisions, human factors, requirements, backlogs, and validation. This is an instruction toolkit, not a trained model, data connector, or autonomous professional service. No configured connectors, hooks, secrets, network services or installers are bundled. An optional local Python renderer produces the progress map HTML companion. The workplace packages are separate and unchanged.

Start here

  1. Read ROUTER.md and choose one relevant task.
  2. Copy the context starters to your project, fill them as needed and explicitly supply them. See CONFIGURATION.md; profiles are not automatically loaded.
  3. Read the selected skills//SKILL.md and supply actual source material.
  4. Ask for the artifact and a separation of evidence, interpretation, assumptions, recommendations and validation needs.
  5. Review claims against sources before implementation or operational use.

Try examples/intake-notes.md with examples/try-it.md. Compare the response with examples/review-guide.md; do not expect identical wording.

What is included

  • 20 general-practice skills adapted for a domain-neutral public edition.
  • domain-context-adaptation replaces mining-specific context.
  • design-system-context replaces workplace design-system assets and rules.
  • Catalog, router, context profiles, fictional examples and evaluation cases.
  • PEB blank template v1.1 and completed maintenance-planning example v1.0 in examples/.
  • Product Lifecycle Progress Map template, illustrative excerpt and optional HTML renderer under skills/product-lifecycle-progress-map/.
  • Codex, Copilot Agent/Space and portable Claude guidance under adapters/.

Use the instructions as readable Markdown in a supported environment. Load only relevant skills. A platform's model, context access, permissions and tools affect results; these files do not grant additional access or authority.

Formats

Portable: use this folder as a reference library. Supply selected instructions and source material explicitly if the environment cannot discover skills. Codex: see adapters/codex/README.md. Copilot: see adapters/copilot/README.md. Claude: see adapters/claude/README.md. Spaces: see adapters/copilot/SPACE-GUIDANCE.md. Snowflake: use portable guidance as a starting point; no account configuration or validated import adapter is provided in this edition.

Validation status

See VALIDATION.md. File checks establish package structure, not reliable agent behavior or product outcomes. Run the evaluation cases in your target environment before wider adoption and retain its version and model details.

Rights and provenance

This edition adapts the author's HSE practice; company-specific skills, token exports, internal URLs and workplace examples are excluded. See NOTICE.md and LICENSE.md for the current status. No third-party models or standards are bundled.

Research and method attribution

See FOUNDATIONS.md for source authors, applicability, and limitations. This separates underlying research from the original toolkit and its citation.

How skills are selected

Once installed and discovered in Codex, relevant skills can be selected automatically from your request. You can also name a skill directly when you want a particular method. Selection depends on the environment and task; verify both explicit and automatic selection during setup. To check automatic selection, describe a representative task in a fresh chat without naming the skill, then inspect which instructions the assistant used.

Other environments may require selecting an agent or supplying instructions explicitly. Follow the adapter guide for your environment rather than assuming Codex discovery behavior applies everywhere.