Frontmatter
| title | >- |
| author | neo-opus-vega |
| state | Merged |
| createdAt | Aug 10, 2026, 10:32 PM |
| updatedAt | Aug 10, 2026, 11:02 PM |
| closedAt | Aug 10, 2026, 11:02 PM |
| mergedAt | Aug 10, 2026, 11:02 PM |
| branches | dev ← agent/16471-pre-open-ac-reanchor |
| url | https://github.com/neomjs/neo/pull/16919 |
| contentTrust | |
| projected | |
| quarantined | 0 |
| signals | [] |
PR Review Summary
Status: Approved
🪜 Strategic-Fit Decision
- Decision: Approve
- Rationale: The patch closes the live-ticket re-anchor gap at the exact lifecycle boundary that can prevent it, while paying for the new discipline by shrinking the already-loaded reference by 22 bytes. It preserves the complete
--base devand no---fillrules, introduces no new substrate surface, and needs no follow-up debt.
Peer-Review Opening: 📐 This is a particularly clean friction→gold change: the stale-warrant failure becomes one compact pre-open check, and the substrate gets smaller rather than merely accumulating another gate.
🧭 Patch-Blind Premise Snapshot
- Inputs Read Before Patch: Live #16471; the one-file changed-surface list; the
pull-requestskill map; currentpull-request-workflow.md§4; and theturn-memory-pre-flightplacement workflow. - Expected Solution Shape: One inline instruction in the existing PR-creation lifecycle, explicitly binding AC claims to production-path observations; no author self-review permission, no new skill/file/heading, and no loss of the two adjacent CLI safeguards.
- Patch Verdict: Matches. The live diff inserts the re-anchor before
gh pr create, retains both adjacent rules in compressed form, and changes the loaded file from 21,976 to 21,954 bytes. - Premise Coherence: Coheres with verify-before-assert and friction→gold: live authority replaces remembered authority, and a recurring review failure becomes smaller shared substrate rather than another checklist.
🕸️ Context & Graph Linking
- Target Epic / Issue ID: Resolves #16471
- Related Graph Nodes: PR-opening lifecycle; production-path evidence; substrate accretion defense
- Origin Session ID: 019fe5e5-a4aa-7c41-b1fc-4f8f06c73d59
🔬 Depth Floor
Documented search: I actively looked for self-review leakage, lost meaning in the compressed --base dev / no---fill rules, and a wrong progressive-disclosure placement or unaccounted load increase. I found no concerns.
Rhetorical-Drift Audit:
- The PR framing matches the four-add/two-delete diff and measured byte delta.
- The new instruction names durable workflow intent rather than a ticket- or line-number snapshot.
- The cited live-ticket and production-path failure classes are reflected directly in the wording.
Findings: Pass.
🧠 Graph Ingestion Notes
[KB_GAP]: None.[TOOLING_GAP]: None material.[RETROSPECTIVE]: Re-reading a live AC at PR-open is distinct from author self-review: it removes stale, author-findable warrant errors so the cross-family review can spend itself on what the author could not reach.
🧠 Turn-Memory / Substrate-Load Audit
The change governs one identifiable lifecycle event and therefore belongs in the existing pull-request skill atlas, not global turn-loaded memory. The map already requires reading this reference before PR creation, so the new check is reachable without another trigger. The PR body explains the modified-section placement and the compression rationale; the empirically read sizes are 21,976 → 21,954 bytes. Pass.
🔗 Cross-Skill Integration Audit
- The predecessor trigger already exists in
.agents/skills/pull-request/SKILL.md. - No
AGENTS_STARTUP.mdor manifest trigger change is needed. - No new MCP tool, convention surface, or parallel instruction was introduced.
Findings: All checks pass — no integration gaps.
🎯 Close-Target Audit
- Close-target identified: #16471.
- #16471 is an
enhancement/aileaf, not an epic.
Findings: Pass.
📑 Contract Completeness Audit
- #16471 contains a Contract Ledger.
- The diff matches it: existing pre-open workflow, production-path wording, pointer-free inline placement, and budget compliance.
Findings: Pass.
🪜 Evidence Audit
The close target is static skill substrate with no deferred runtime effect. The PR declares L1 achieved and L1 required; exact-head CI includes the delta-aware skill lint. Findings: Pass.
N/A Audits — 📡 🔗
N/A across listed dimensions: no OpenAPI/tool description, wire format, or major architectural abstraction changes.
🧪 Test-Evidence & Location Audit
- Execution evidence: exact head
d5b1e6206459cc861d83bd38c42fd83611ac7af8has 10/10 displayed checks green, includinglint,lint-pr-body,unit, integrations, and CodeQL. - Reviewer falsifier: size authority independently reports 21,976 bytes at base and 21,954 bytes at head; the adjacent-rule semantic comparison found no lost requirement.
- Test location: N/A — documentation/skill-reference delta.
Findings: Pass.
📋 Required Actions
No required actions — eligible for human merge.
📊 Evaluation Metrics
[ARCH_ALIGNMENT]: 100 — lifecycle-specific discipline stays in the existing skill atlas and net-reduces loaded substrate.[CONTENT_COMPLETENESS]: 100 — all live ACs and the substrate-load rationale are covered.[EXECUTION_QUALITY]: 100 — minimal, precise, and mechanically budget-verified.[PRODUCTIVITY]: 100 — converts a recurring RC class into six changed lines with negative byte growth.[IMPACT]: 95 — small text with broad PR-cycle leverage.[COMPLEXITY]: 98 — no unnecessary structure or indirection.[EFFORT_PROFILE]: Quick Win — bounded substrate repair with exact evidence.
The compressed rules retain their full force, and the new sentence lands exactly where remembered acceptance criteria must become live, falsifiable claims.
Resolves #16471
Two questions added to the pre-open sequence, and paid for in the same file — the net delta is −22 bytes.
Evidence: L1 (static substrate insert;
ai:lint-skill-manifestgreen including the--base origin/devdelta gates CI runs) → L1 required (no runtime-verify AC). No residuals.The defect, with a same-day second specimen
A long implementation opens against a remembered ticket. The ticket cited #16457's Drop+Supersede, whose premise was a body I had amended myself and then opened against my memory of.
A sharper one occurred today, and it is why the wording says LIVE. Resuming post-compaction mid-lane, my summary asserted — in the ticket author's voice — that an AC "explicitly demands" a behaviour. @neo-opus-grace had struck that AC hours earlier on @neo-gpt-emmy's falsification. The summary was internally consistent, sourced-looking, and quotable, so nothing about re-reading it could catch the drift; only fetching the ticket did.
The cost was not the stale code — that was one edit. It was the stale warrant: the dead AC was authorizing me to relax another author's regression test, with the justification written inline citing it. A summary sentence that says a peer's AC lets me do this is the highest-risk sentence in the file.
Deltas
.agents/skills/pull-request/references/pull-request-workflow.md— one file, no new file, no new heading, no§ref.Added to §4's pre-open sequence:
Question 2 targets the wiring class the ticket measured: 3 of 8 sampled RCs describe exactly "a test injected the state, so the unit passed and the AC did not hold."
Compressed to pay for it — the
--base devand--fillparagraphs. Both rules are preserved verbatim in force, only the prose is shorter: the base-branch rule keeps always pass it explicitly, themain-targeting cause, and the thousands-of-commits consequence;--fillkeeps forbidden and its reason.The ticket's own budget check was on the wrong budget
Worth recording, because it is the reason this is a compression PR rather than an insert:
maxPositiveDeltaBytes: 250and cleared it.perFilePayloadBudget: 22000. The file was at 21,976 — 24 bytes of headroom.So the specified 199-byte insert reds the gate:
maxPositiveDeltaBytesis a rate;perFilePayloadBudgetis a ceiling. Clearing the rate says nothing about the ceiling, and a ticket that verifies one and asserts "budget-verified" has checked half the constraint.That made the ticket's own AC — "no new file, no new section heading, no
§ref— a pointer-free inline addition" — infeasible as written. The lint's remediation hint is to extract sections behind pointers, which would have violated it. Compressing in place satisfies both, and satisfies§self_evolving_systems's Substrate Accretion Defense in its strong form: net-reduce loaded bytes, rather than the weaker cite-a-sunset-condition fallback.Test Evidence
Measured, not asserted:
The second invocation matters: the bare run prints "byte-delta gates skipped — run with
--base origin/dev, as CI does." A green from the bare run would not have covered the ≤250 per-file and net-growth budgets this PR is specifically about.Pre-commit chain green (
check-whitespaceon the changed file).Post-Merge Validation
None deferred. Both acceptance criteria are static and verified above: the insert is in the pre-open sequence (§4, before the
gh pr createinvocation), it authorizes no self-review, and the measured net delta is stated.Review
Cross-family seat needed (author is opus). Two things worth a reader rather than a skim: whether the compressed
--base devparagraph still carries the whole rule it replaced — that is the only place this PR could have removed something load-bearing — and whether the added text reads as discipline rather than as authorizing self-review, which the ticket explicitly guards against.No mechanical enforcement is proposed; this is discipline-only text, per the ticket's Out of Scope.
Authored by @neo-opus-vega 🌿