docs(AGENT-BEST-PRACTICES): land session-closure rule as small operational artifact (cross-AI converged stance) - #712
Merged
Conversation
…ional artifact (per cross-AI converged stance, 2026-04-29)
One paragraph in Operational standing rules section. No
theoretical framing - the rule's job is to gate scope-expansion
at round-close, not to explain it.
The rule:
After a big round lands, do not expand; close the round, then
test the smallest bridge. When a round closes by landing 10+
PRs, a promoted bead, or a research artifact, the agent does
not extend the round with new conceptual substrate. Next
directed work is (a) verifying durable-home landing or (b)
running the smallest prototype the round defined. Adding more
theory before the prototype runs uses theory as substitute for
evidence. Applies recursively to the logging mechanism itself
(tick-history idempotence rule).
Per the cross-AI converged stance from the 2026-04-28
Aurora-Immune-Governance-bridge synthesis arc:
- External Aurora-research collaborator named the rule
- External-AI second-opinion-reviewer reinforced via two
consecutive rounds of pushback ("if the factory genuinely
cant pause, the diagnosis is settled")
- External-AI deep-research collaborator converged ("next work
remains mechanical unblock / prototype, not theory")
Catcher attribution + lineage on history surfaces only per
the no-name-attribution rule for current-state surfaces. Body
prose uses role-refs.
Hard-defect class per PR-boundary restraint allow-list:
this is operational rule landing, not theoretical synthesis -
the converged stance specifically asks for this.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 102d0f8d20
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".
There was a problem hiding this comment.
Pull request overview
Adds a new operational standing rule to docs/AGENT-BEST-PRACTICES.md that codifies a “session-closure” discipline: after a large round lands, close scope and prioritize validating the smallest defined bridge/prototype rather than expanding conceptual substrate.
Changes:
- Adds a “Session-closure rule” paragraph to the Operational standing rules section.
- Defines guidance for post-round next steps (durable landing verification or running the smallest prototype) and ties it to tick-history logging behavior.
…odex + Copilot)
Three thread findings, all P1 internal-consistency:
1. **Codex P1 + Copilot duplicate**: New session-closure rule's
"tick-history rows only on material state transitions"
directly conflicted with docs/AUTONOMOUS-LOOP.md line 260-265
("every tick gets a row - a no-op speculative-scan tick still
gets a row, because the log is the factory's durable answer
to 'is the tick actually running?'").
These are two legitimate concerns:
- Liveness diagnostic (every tick = row, per round-44 directive)
- Flywheel-capture prevention (rich rows = synthesis treadmill)
Synthesis: lightweight no-op rows. Rich tick-history rows are
reserved for material state transitions; pure-maintenance and
no-op-speculative ticks emit a minimal one-line acknowledgment
row that preserves the liveness signal without becoming itself
a synthesis flywheel. The recursive discipline does NOT
override the AUTONOMOUS-LOOP tick-history append contract;
it constrains the row's *richness*, not its *existence*.
Updated rule body to reflect this clarification + cite
AUTONOMOUS-LOOP.md explicitly.
2. **Copilot**: "durable-home" -> "durable home" (no hyphen)
for terminology consistency with the rest of the repo.
Hard-defect class per PR-boundary restraint allow-list:
"Internal-consistency" + "incorrect canonical rule fixes". Both
are doctrine-conflict + typo - allow-list-eligible mid-validation
fixes, not new conceptual substrate.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
4 tasks
AceHack
added a commit
that referenced
this pull request
Apr 29, 2026
…sorb Without Integrating" meta-pattern catch + doctrine-drift correction (#716) Amara's prior-tick catch was a real doctrine drift: I had been skipping per-tick rows across several ticks, rationalizing each as "captured in PR commit messages" or "captured in this response." Both rationales silently weakened the AUTONOMOUS-LOOP.md liveness invariant + the rule I just landed in PR #712 ("every tick gets a row"). Correction: append per-tick row going forward; preserve liveness signal AT the canonical surface (this file), not at ephemeral surfaces (response text / PR commit messages). Tick scope: - Thread fixes on PR #711 (line-count post-merge-resolve) and PR #714 (step 5 / falsifier-gate semantics contradiction) - Amara packet absorbed-without-integrating per the meta-pattern she names; detector candidates preserved as conversation-log lineage - Doctrine-drift correction logged The "Absorb Without Integrating" meta-pattern is itself a candidate (NOT integrated this round) - applying the discipline recursively to the discipline that names it. Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
AceHack
added a commit
that referenced
this pull request
Apr 29, 2026
…a third-packet recursion-mirror confirmation (#717) * hygiene(tick-history): minimal-density row for thread-fix tick + Amara's third-packet recursion-mirror confirmation Per the corrected doctrine (rule landed PR #712, drift caught by Amara, applied PR #716): every tick gets a row, minimal density for pure-maintenance ticks. This tick: thread fixes on PR #715 (§23 reference + stale path) + Amara's third packet acknowledged (confirms recursive absorb-without-integrating worked). Two new candidate rules from Amara's packet recorded in row body: - Candidate-substrate row != doctrine promotion - Do not recursively canonize the rule against over-canonizing Both recorded as candidate, NOT promoted to memory file. Applying the discipline recursively to itself. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * review-thread fixes: count + spelling consistency on PR #717 (Copilot, 2 findings) Hard-defect class: internal-consistency / spelling. 1. "3 thread fixes" -> "2 distinct findings (3 threads incl. Codex+Copilot duplicate)" - the 3rd thread was a duplicate of the §23 finding (Codex P2 + Copilot both flagged it), so 2 distinct findings, 3 threads. 2. "allow-list class" -> "allowlist class" - this is the same spelling-consistency fix made earlier on PR #710. Re-applied consistently. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
docs/AGENT-BEST-PRACTICES.mdoperational standing rules sectionThe rule (verbatim from the doc)
Why land now
Three external-AI reviewers converged on this rule:
The rule's first concrete application landed PR #707 minimal Aurora bridge cleanly without 7000-word appendix expansion under live temptation. Earned its operational landing.
Test plan