skill: software-delivery-explainer — where agents act and where humans keep control

[wz-gsa]

Part of #319 (Executive Technical Explainer epic).

What

A communications skill that explains how software moves through a modern engineering process and, crucially, where agents participate and where the human control points are.

Baseline model (adaptable, not one-size-fits-all):
need/problem → issue → isolated branch/workspace → implementation → automated tests → pull request → CI/security checks → human review → merge → deployment.

Central honest thesis

The agent accelerates work inside the engineering process; it does not inherently replace the process or its controls. It performs more work between the controls.

Make human/automated control points explicit. Show which portions can be delegated and which remain consequential human decisions.

Ground truth (cite, don't invent)

• Human approval gates + no self-approve/self-merge: playbook AGENTS.md §3.2/§8.2/§14.2.
• Separation of duties: SECURITY-CONTROLS.md AC-5. Human author-of-record: AI-CONTRIBUTION-POLICY.md.
• CI/CD stages: playbook docs/GETTING-STARTED.md §7.
• Avoid implying every org uses the identical workflow.

Acceptance

• make validate + make generate clean.
• tests/test-cases.yml: asserts human review/merge control points are named; asserts the output does NOT state the agent merges or approves its own work; "show where humans remain accountable" prompt.

AI-assisted (OpenCode).

[wz-gsa]

Implemented and merged in #329 (verified on main: the three communications skills, the concept-library/audience-profiles/measurement-guidance reference, the technical-explainer design-artifact profile + taxonomy entry, evals, and the 7/7 adversarial review are all present). Closing as completed.