review-7.md
c133 · Phase 7 review stop: the first training on a topic (2026-11-12)
Built in the c142 session right after Mark ruled Phase 6's calls (D45–D51, review-6.md § Ruled), at his call 5d: "Yes: Observe and Enforce".
Where to look
| What | Where | Why |
|---|---|---|
| The library, four trainings | /pages/phase4/mock/training-library | Observe and Enforce, "a topic", beside the three policy trainings |
| The topic, taken on its own | step 1 · step 3 · the record | four steps, four questions, four new scenes; the record names the topic |
| The preview page | /pages/phase4/trainings | the Policy or topic column; the scene gallery draws the topic's scenes as a reader takes it |
| The count | Compliance, took=T-MODES-1,T-SECRETS-1 | a topic and a policy's training counted the same way |
Captures: thread "Phase 7 · the first training on a topic…" (scripts/shot-programs/c133-p7.json, five shots).
What was done
- The model (D49).
Training.policyIdis optional besidetopic: { id, label }, exactly one.TrainingRecordnamestopicwhen there's no policy. The parser refuses a file with both, with neither, or with a malformed topic. A topic training's# Assignmentneeds onlymatters:. Policy trainings serialise as before, so their hashes didn't move.findingsForTraining,POLICY_TRAININGS, the library rows and slots, and the API list all handle a training with no policy: the library fills a topic from the reader alone and IT answers its questions. - Observe and Enforce (T-MODES-1,
en/observe-and-enforce.md), 2 minutes. The topic is three of the guide's concepts: Observe, Enforce and Off.- The steps: what a policy does on your machine (three modes, set per team); Observe (it goes through, a finding records it, a record and not a verdict); Enforce (stopped on the machine where the policy can stop it, and some policies can only record); what to do when a policy stops your work (
instead: ask IT or the owner, sanctioned or an exception; until then, sanctioned tools). - The check: four questions, with the right answers second, first, third and second (D34).
- The scenes: four new ones from the pawns and the sketch kit,
policy-modes,mode-observe(the flow),mode-enforceandmode-ask, catalogued and mapped. - Voice: every sentence through
check-brand-voice(productUI), zero violations.
- The steps: what a policy does on your machine (three modes, set per team); Observe (it goes through, a finding records it, a record and not a verdict); Enforce (stopped on the machine where the policy can stop it, and some policies can only record); what to do when a policy stops your work (
- The pages.
- The library lists it as "a topic: Observe and Enforce" and opens it "on the topic". The contact line drops "or the policy".
- The record card reads "no finding; the topic is observe-and-enforce".
- The preview page has a Policy or topic column, a "take it on its own" link where a topic has no finding, and "Written for anyone" in its On its own line. The gallery draws the four new scenes as User A reads them, linked to the library step: 16 scenes, none blank.
- The Compliance report counts it with the rest.
- The docs:
authoring.md(the topic fields and refusals),publish.md(topicin the JSON),handoff.md(a new § 7, Trainings on a topic, for c134), and the/trainingskill (a section for topic trainings).
Checked on :3000 in the browser pane: the API publishes four trainings and withholds none; the preview page shows four, all voice passes, and 16 scenes drawn; the library row, lesson and record for T-MODES-1; the Compliance count with took=T-MODES-1,T-SECRETS-1. No failed requests.
The calls for you
- The id:
T-MODES-1, named for what it covers. It follows theT-<HAZARD>-<n>pattern (c137 E7), with the topic in place of the hazard. The alternative isT-TOPIC-<n>, which sorts topics together but says less. Proposed: keep. - One training for three concepts. Observe, Enforce and Off are three of the guide's 18 concepts, but they only make sense together. Proposed: a topic may cover several concepts where one doesn't stand alone; the seed list (D50) becomes a list of trainings, not one per concept.
- "A finding is a record, not a verdict." Step 2 says it to someone who may have just had their first finding. Proposed: keep. It's the register R6 asks for (a safety briefing, not a reprimand).
- The claim in step 3 that some policies can only record ("a sign-in to a personal account happens outside qcontrol's reach") repeats T-ACCT-1's question. Proposed: keep, and check it with Tyler with D22–D27.
- The operator's send (D51) isn't mocked: a topic has no finding to start from, so it needs a place in the operator's UI to pick a topic and a person. Proposed: c134's, recorded in the hand-off.
Next topics (D50)
From the guide's list, the ones a person meets first: Finding and Severity, Sanctioned, Exception and Scope, and Endpoint and Harness. The operator terms (Draft, Publish, Drill, Connector, Source checks, Framework, Enforcement Posture, Spend ceiling) are for operators taking them on their own. Each is a file through the /training skill's topic section.
Ruled (2026-11-12)
Mark, the same day: "I've read and agree with your defaults, go with those." All five as proposed, D52–D56 (plan.md). Topic ids are named for what they cover (D52). A topic may cover several concepts (D53). "A record, not a verdict" stays (D54). The claim that some policies only record stays, to check with Tyler (D55). The operator's send is c134's (D56).