Skip to content

docs(memory): mark the checkpoints closed and correct main HEAD - #688

Closed
adamgell wants to merge 3 commits into
mainfrom
docs/memory-checkpoint-currency
Closed

adamgell wants to merge 3 commits into
mainfrom
docs/memory-checkpoint-currency

Conversation

@adamgell

Copy link
Copy Markdown
Owner

Documentation only — no behaviour change.

The problem is worse than staleness

memory.md is described as "durable facts … loaded every turn for this project", and it presented four checkpoints as live lanes:

Issue Claimed state Actual
#320 client health "6/6 focused pass … NOT merge-ready" CLOSED 2026-08-05
#329 DP lifecycle "P1 open … hold PR until clean" CLOSED 2026-08-05
#330 SUP coverage "Full gate pending: …" CLOSED 2026-08-05
#366 Intune CP "Findings: observedThroughLine must dominate …" CLOSED 2026-08-08

All four closed as implementation history, and the closing notes say the blockers were scoped out rather than resolved. #320's says it is "intentionally not merge-ready" and claims no native Windows acceptance; #366's closeout states that closure "is not final epic acceptance evidence".

So the table did not merely describe the past — it invited the wrong action. An agent reading it would try to clear blockers on work that was deliberately closed, and #329's row in particular reads as an instruction: "hold PR until clean".

Main HEAD was also named as a9a67422 from 2026-08-03; it is 1967a0a1 from today.

Change

  • Main HEAD corrected.
  • Each checkpoint row now records its closing date and what closure meant, sourced from the issues' own closing comments rather than inferred.
  • The standing gap — native Windows acceptance, which none of these checkpoints claims — is stated once instead of being implied by four blocker lists.
  • Recovery branch table untouched; all six branches still exist.

One thing this PR does not fix

The same four-row table appears in .omp/AGENTS.md under "Verified Checkpoints (From PM Charter)", with the same stale states. I have not touched that file — it is the operator's context surface rather than repository documentation, and editing it is not mine to do. Whoever owns it should mirror this change.

memory.md presented four checkpoints (#320, #329, #330, #366) as live
lanes with open blockers, and named main HEAD as `a9a67422` from
2026-08-03. All four issues closed in August 2026 as implementation
history, and main is at `1967a0a1`.

The stale table was not merely out of date, it invited the wrong action:
an agent reading it would try to unblock work that was deliberately
scoped out. The closing notes say so explicitly - #320 is "intentionally
not merge-ready" with no native Windows acceptance claimed, and #366's
closeout states that closure is not final epic acceptance evidence.

Each row now records the closing date and what closure actually meant,
and the standing gap - native Windows acceptance - is stated once rather
than implied by four blocker lists.

The recovery branch table is untouched; all six branches still exist.
@coderabbitai

coderabbitai Bot commented Sep 24, 2026 •

Copy link
Copy Markdown

Warning

Review limit reached

Next included review available in 28 minutes.

Check out review usage here.

View limit details

Limit details: You’ve used all 4 included reviews currently available. Your 66 included PR review attempts over the past 7 days set your current allowance at 4 reviews per hour.

Your organization has reached its usage spending cap. Adjust your spending cap in the billing tab.

Learn how review limits work.

Review configuration:

⚙️ Run configuration

Configuration used: Repository: adamgell/cmtraceopen/.coderabbit.yaml

Review profile: ASSERTIVE

Plan: Team

Run ID: d69f4428-e5a6-4525-b9ed-c670ff092628

📥 Commits

Reviewing files that changed from the base of the PR and between e3410c8 and feae72b.

📒 Files selected for processing (2)
  • memory.md
  • soul.md

Comment @coderabbitai help to get the list of available commands.

The same four-row checkpoint table appears here as in memory.md, with
the same stale states - all four issues closed in August 2026 as
implementation history, with their recorded blockers scoped out rather
than resolved.

`soul.md` is the identity file, so an agent reading "NOT merge-ready"
and "Hold PR until clean" here has the strongest possible reason to
believe those lanes are live and need clearing.

Rows now carry the closing date and what closure meant, and the standing
gap - native Windows acceptance - is stated once.
memory.md told a reader that Theo's docs-audit phases live at
`docs/audit-phase2` and `docs/audit-phase3`. Neither directory exists,
and neither was ever committed: git log --all --diff-filter=A on those
paths returns nothing.

The note that the audits happened is worth keeping; the locations are
not, because they send a reader looking for something that is not there.
@adamgell

Copy link
Copy Markdown
Owner Author

Independent verification at this head. The table's facts all hold; one attribution does not, and for a file described as durable facts that is worth one line.

Verified:

Claim Check
all four issues closed CLOSED - #320/#329/#330 at 2026-08-05, #366 at 2026-08-08
the dates in the table exactly those
every SHA written all resolve to real commits in this repository (6ccf8dafa791ad7d07d3b7bb450e6fe31e8dfb3c, a03af515fa…, 76e2b0b910d…, 04e1ecba6f…, 1967a0a1)

That last row is the one I most expected to fail - a checkpoint SHA that resolves to nothing is exactly the failure mode of a table like this - and every one of them is a real object.

One attribution to correct. The table says of #320:

The closing note says "intentionally not merge-ready" and claims no native Windows acceptance.

The words are real, but they are not in a closing note. #320 has 29 comments, the last on 2026-08-03T22:46:41Z, and the issue closed on 2026-08-05T00:47:30Z - two days later, with no comment. The quoted text is from that earlier checkpoint publication, which also ends "This issue remains open."

The distinction matters for a fact table: "the project recorded this at closure" and "the issue closed silently and the decision is inferred afterwards" are different statements, and the second is the true one. Suggest something like "the checkpoint comment of 2026-08-03 says … ; the issue closed two days later with no resolution note", and the summary line becomes "closed without resolution notes; the blockers are recorded as scoped out rather than resolved".

Everything else in the table is accurate, including the part that matters most: that a closed checkpoint is not a merged one.

@adamgell

Copy link
Copy Markdown
Owner Author

These three PRs edit the same small set of files, and two pairs of them collide. Worth knowing before one is merged, because merging any one makes the others conflict.

#688 and #679 both edit the same table. Both touch memory.md and soul.md, and in soul.md they hit the identical region:

#679  @@ -127,7 +127,9 @@ These are documented, not speculative. Always verify state before acting:
#688  @@ -127,7 +127,12 @@ These are documented, not speculative. Always verify state before acting:

That is the verified-checkpoints table. In memory.md their hunks sit adjacent (~50 and ~76 for #679, ~21, ~52, ~96 for #688), so they will collide rather than compose.

Recommendation: #688 as the survivor on that overlap, because its claims are the ones I checked - the four issues are closed on exactly the dates it states, and every SHA it writes resolves to a real commit. #679's unique parts are worth carrying across: its library.md note and the ADR-005 line.

#679 and #696 both edit library.md's opening region, and they take opposite positions on the same question:

Deleting without that note removes the routes and the reason; the note without the deletion leaves thirteen dead entries above a paragraph saying they are dead. They belong together, and whoever lands second should reconcile rather than drop the other.

I have not pushed or closed anything here - which survives each overlap is Adam's call.

@adamgell

Copy link
Copy Markdown
Owner Author

#679 fixes the same thing — same two files, same four rows — so worth picking one.

memory.md and soul.md both carry the checkpoints table, and both PRs rewrite it:

#679 #688
review reviewed, 3 findings open REVIEW_REQUIRED, never reviewed
rows corrected #320, #329, #330, #366 the same four
state BLOCKED/CHANGES_REQUESTED BLOCKED/REVIEW_REQUIRED

They do not approach the correction the same way, and that matters. #679 credits a landing PR per row — "MERGED via #340" — and that is where its open findings are: #340 merged into codex/parser-family-skeleton and closed nothing, while #490 merged to main on 2026-08-05 and closed #320, #329 and #330. Its second finding is that "months newer" should be "weeks newer" for an August 30 date against August 4–5 plan links.

This PR's table says CLOSED with the date and no landing-PR citation, which sidesteps the first finding entirely. Neither is a defect — but if #679 is the one that lands, it needs the #340 → #490 correction, and this one does not.

Whichever lands, the other should close, and the corrected state should not be merged twice into the same table.

@adamgell

Copy link
Copy Markdown
Owner Author

Closing as a duplicate of #679, which corrects the same checkpoint tables in memory.md and soul.md. The branch stays in place.

@adamgell adamgell closed this Sep 25, 2026
@adamgell
adamgell deleted the docs/memory-checkpoint-currency branch September 25, 2026 13:51
adamgell added a commit that referenced this pull request Sep 30, 2026
…#735)

Theo's index routed to three docs-audit notes and Roger's to six SCCM
notes. Neither staff directory holds anything but its own index, and none
of the nine paths appears in any commit on any ref, so every route led a
reader to nothing - with no way to tell "not written yet" from "written
somewhere else".

Roger's index keeps the empty-state line Theo's already carries, so an
index with no routes says so rather than looking incomplete.

memory.md's docs-audit line is left alone: PR #688 rewrites it.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

enhancement New feature or request

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant