docs: correct two header claims the last two merges falsified

[os-sam]

Closes #72
Closes #73

Two living statements in source that are now demonstrably false. Both are the kind that stop the next person from finding a bug, not merely the kind that misinform — which is why they are worth a PR rather than a backlog entry.

Comment-only. No metadata, behaviour or translatable string changes.

#72 — src/data/shared/build.ts was wrong in a more interesting way than the card says

The header read:

because a seed row does not go through the stamp hooks (skipTriggers), and a seed that leaned on them would break silently the day a hook changed

The card that filed this reported the flag as unfounded — "skipTriggers appears only in that comment" — which is true of this repository but not of the platform. The flag is real, the seed loader does carry it, and it is the reason the sentence is false rather than evidence that it is:

@objectstack/metadata-protocol, src/seed-loader.ts, the SEED_OPTIONS docblock:

skipTriggers suppresses record-change AUTOMATION (autolaunched flow triggers) for seed writes: a package's seed is pre-existing END-STATE reference/sample data, not a stream of user events … Lifecycle HOOKS (derived/default fields, validation) still run.

So the loader documents exactly the distinction this comment collapsed, one sentence later, and the answer is the opposite of what the comment asserted. Every stamp hook in src/hooks/ fires on seeded rows.

That is how #65 happened. An unconditional last_activity_at assignment overwrote the authored value on all 200 applications, and a header stating that hooks cannot reach a seed row is precisely what would have stopped someone from looking there. A false comment that closes off a line of enquiry costs more than one that merely misleads.

The rewrite states what the loader actually does, keeps the real reason the seed denormalises explicitly (it should state the facts it means rather than inherit whatever a hook computes today), and records the neighbouring flag while it is in view: seedReplay skips the object's state_machine entry and transition checks, which is why an ats_offer can be seeded straight into its terminal accepted state — the thing #53 relies on and #64's review verified empirically (818 inserted, errored: 0) without knowing the mechanism.

#73 — F4's header, falsified by #65 four hours later

src/flows/application-stage-notify.flow.ts read:

stamps last_activity_at = now() on every application update, unconditionally (stamp.hook.ts, last line of the handler)

True when written; false since #65. The hook now stamps on any update whose payload names a field of the application.

The paragraph's conclusion is unchanged — a stage change is such a payload, so the value is already correct by the time this record-after-update flow starts, and the flow still deliberately has no second writer. Only the reason moved, and it now points at the rule rather than at a line that no longer exists.

Gates

validate=0   lint=0   typecheck=0
i18n source parity: 512 keys · 512 compared · ✓ every bundle key restates the label its metadata declares

🤖 Generated with Claude Code

https://claude.ai/code/session_01PbJ5Cy9KDAzeQHo8bsMadG

---

Generated by Claude Code