← Documentation home

Canonical Markdown source · Oct 20, 2018

Evidence Script Ownership

ops/evidence-script-ownership.md · 73 lines · SHA-256 33e097b991f8

The readiness scripts are part of the product, not a parallel app. Every

high-churn evidence command has one accountable owner lane, one artifact surface,

and at least one package-script check that must stay valid when the script

changes.

The machine-readable source of truth is

`src/services/evidence-script-ownership.ts`. The guard is

`tests/services/evidence-script-ownership.test.ts`, which fails when a

high-churn package command loses its owner, points at a missing script, or names

a non-existent required check. It also fails when a colon-delimited

`package.json` script namespace such as `launch:`, `activity:`, `pilot:*`, or

`monitoring:*` lacks a namespace owner, preferred entry points, and pruning

guidance.

Owner Lanes

| Owner lane | Scope | Default cadence |

|---|---|---|

| Strict readiness owner | `pnpm review:goals*`, closeout, and the local-vs-strict claim boundary | Per change |

| Launch readiness owner | preflight, launch review, launch evidence, a11y, crawler, public trust, and smoke packet commands | Per launch |

| Long-term evidence owner | public k6, telemetry, Era C exit gate, 30-day runway, and maintenance evidence | Weekly |

| ActivityStreams adoption owner | external consumer, durable subscription guard, adoption, and partner handoff evidence | Weekly |

| Managed pilot owner | outreach, buyer pack, activation, support, KPI, packaging, and paid-pilot proof | Weekly |

| Continuity hardening owner | DR, pentest, disabled-system staging drill, and projection readiness evidence | Weekly |

| AI eval owner | golden-question evals, AG2 worker evals, and AI baseline artifacts | Per change |

| Worker operations owner | outbox, publication queue, supervised Facebook Page publishing and campaign packaging, projection drain, dead-letter, and alert commands | Weekly |

| Product governance owner | privacy, product constraints, research A+ readiness, credibility, validation drift, docs stats, and conformance proof | Per change |

Change Rule

When adding or changing any high-churn evidence command:

  1. Update `src/services/evidence-script-ownership.ts` with the package alias,

backing script, artifact path, required check, owner lane, and review cadence.

  1. Update the matching runbook or artifact contract doc when operator behavior

changes.

  1. Run

`node --import tsx --test tests/services/evidence-script-ownership.test.ts`.

  1. Preserve the strict-readiness distinction: local checks may prove the

machinery is healthy, but partner, customer, uptime, SLO, and paid-pilot

proof must remain real-world evidence until observed.

Escalation

If ownership is unclear, assign the command to the lane that consumes the

artifact in `/readiness` or `pnpm review:goals`. If two lanes consume it, choose

the lane that can verify the artifact without enabling external side effects,

then document the dependent lane in the purpose text.

No command should be owned only by a README paragraph, CI workflow, or one-off

operator memory. The registry is the durable handoff.

Namespace Grouping

Package-script prefixes are treated as operator-facing command groups. When

adding a new `namespace:command` alias, keep it inside an existing namespace

when possible. If a new namespace is necessary, add it to

`PACKAGE_SCRIPT_NAMESPACE_CONTROLS` with:

  • a registered owner lane
  • the preferred entry point commands maintainers should try first
  • a pruning rule that explains when variants should be collapsed or removed

The namespace registry is the source of truth for the current count; every

namespace, including research automation, must remain classified with no

unowned aliases.

`content:*` belongs to the AI eval owner. Its preferred entry point is

`content:generate:trial`; additional drafting aliases must reuse that governed

trial contract unless they produce a distinct retained evaluation artifact.