docs(ignition): restart the acceptance clock at the 15-min schedule flip

[kyle-sexton]

Follow-up to #1052, which merged while its final review thread (PLAN.md L285, restart the acceptance window after the cadence flip) was still open. This lands the fix that thread asked for: the roadmap's >=1-week fires-on-schedule acceptance stretch cannot carry the hourly portion across the cadence flip — hourly fires miss three of every four slots under the 15-min grid — so the acceptance clock restarts at the actual Desktop schedule-flip timestamp. Recorded as the same predicate-vs-acceptance consequence split the 2026-07-21 pause incident established (predicate window unaffected; it measures mature completions, not slot coverage).

Related

• docs(ignition): predicate re-bind ratification + cadence + delegated-merge amendment #1052 (merged; thread docs(ignition): predicate re-bind ratification + cadence + delegated-merge amendment #1052 (comment) remains the source finding). No linked issue — status-note documentation; nothing in this repo closes.

🤖 Generated with Claude Code

https://claude.ai/code/session_017P1vVA8iViUTfQWjA9tgZG

[claude]

Claude finished @kyle-sexton's task in 51s —— View job

---

Review

• Read changed file and diff
• Analyze correctness and consistency
• Post review

---

Summary: Documentation-only change to docs/topics/autonomy-ignition/PLAN.md closing an outstanding review thread from #1052. The added prose records that the roadmap acceptance clock restarts at the schedule-flip timestamp, and that the C2 predicate window is unaffected. No code or config surface changed.

---

Correctness — LGTM

The core reasoning holds: under a 15-min slot grid, an hourly cadence fires in only 1 of every 4 expected slots (~25% coverage). Carrying an hourly acceptance stretch forward as evidence of on-schedule 15-min operation would overstate compliance. Restarting the acceptance clock at the actual schedule-flip timestamp is the correct fix.

The predicate/acceptance split mirrors the 2026-07-21 pause incident precedent (predicate window measures mature completions, so slot-coverage changes don't touch it; acceptance clock measures on-schedule operation, so cadence changes do). Applying the same split here is consistent and well-motivated.

Prose and style — LGTM

The added text matches the surrounding blockquote, uses the established ALLCAPS emphasis convention (PREDICATE, ACCEPTANCE, RESTARTS), and quotes the acceptance criterion verbatim ("fires on schedule with zero manual kicks for ≥1 week") for traceability. Sentence density is high but consistent with the existing section.

One minor observation (non-blocking)

The phrase "the prior hourly stretch misses three of every four slots under the new grid" is accurate (1-of-4 fires → 3-of-4 misses), but a reader skimming quickly might parse it as "the stretch itself is missing slots" rather than "hourly fires cover only 25% of 15-min slots." The parenthetical immediately following clarifies the intent; no change required, just noting it.

Permalink to changed lines:

claude-code-plugins/docs/topics/autonomy-ignition/PLAN.md

Lines 289 to 295 in 8dd0ab7

	> beyond failure tracker items and run history. Same consequence split as the 2026-07-21
	> pause incident: the C2 PREDICATE window is unaffected by the cadence flip (its span
	> measures mature completions, not slot coverage), but the roadmap ACCEPTANCE clock
	> ("fires on schedule with zero manual kicks for ≥1 week") RESTARTS at the timestamp the
	> Desktop task's schedule actually flips to the 15-min grid — the prior hourly stretch
	> misses three of every four slots under the new grid and cannot count toward a week of
	> on-schedule 15-min operation.

---

Verdict: Approve. The logic is sound, the change closes the identified thread, and the prose is consistent with the document's existing conventions.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 8dd0ab755f

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[claude]

Claude finished @kyle-sexton's task in 53s —— View job

---

Review

• Read changed file and full diff
• Assess correctness of Phase 4 sanity-check anchor (new commit bcfc28b)
• Post review

---

Summary: Two commits on this PR. The first (8dd0ab7) adds the acceptance-clock-restart prose at L289–295. The second (bcfc28b) updates the Phase 4 sanity check to explicitly anchor "day 7" to the schedule-flip timestamp — a direct response to the Codex P2 finding.

---

Commit 8dd0ab7 — LGTM (unchanged from prior review)

The acceptance-clock-restart reasoning holds: hourly fires cover 1 of every 4 15-min slots (25% slot coverage), so an hourly accumulation stretch cannot serve as evidence of on-schedule 15-min operation. Restarting the clock at the actual flip timestamp is correct. The predicate/acceptance split mirrors the 2026-07-21 pause incident precedent and is applied consistently.

Commit bcfc28b — LGTM

The Codex P2 finding was: the sanity check said "at day 7" without anchoring to the flip, so a reviewer could execute it against the original accumulation day 7 and inadvertently count the invalid hourly stretch. The new wording closes that gap precisely:

Sanity Check: at day 7 of the acceptance clock — seven days after the recorded timestamp at which the Desktop task's schedule actually flipped to the 15-min grid, NOT day 7 of the original accumulation window

This is as unambiguous as the prose can be. It:

• Names the reference point explicitly ("recorded timestamp… actually flipped")
• Uses capitalized NOT to guard the exact misreading the finding flagged
• Aligns with the restart predicate added in the first commit — they now form a consistent pair: the clock restarts at the flip (L292–293), and the sanity check runs at day 7 of that restarted clock (L341–343)

No gaps left from the Codex finding.

Verdict: Approve. Both commits are correct, the Codex P2 finding is fully addressed, and the prose changes are internally consistent.