choiyounggi · dch0202-rsquare · Aug 13, 2026 · Aug 13, 2026
@@ -1,53 +1,149 @@
-# Knowledge consolidation — 15 open PRs (#17–#40) → one reconciled state
+# Knowledge flush — 3 insight(s)
 
-The 15 open `knowledge/*` PRs (created 2026-08-04 → 2026-08-05, before the
-harvest processed-store dedupe fix in #41) contained 123 file-versions of ~75
-unique pages, with the same insight landing at up to 3 different paths across
-up to 8 PRs. Per-PR review would re-import those duplicates, so — as with the
-#6–#13 consolidation — this branch carries the reconciled end-state and the 15
-PRs are closed in its favor.
+Queue drained: 3 pending candidates from 3 sessions. Outcome: **1 merged into an
+existing page** (this PR), **1 folded into open PR #76**, **1 dropped as a
+same-day duplicate of the merged one**. No new page was created.
 
 ## Verified best-practice
 
-Every adopted page's sources were carried from its originating PR's flush, where
-they were live-verified at flush time; no new URLs were introduced during
-consolidation (checked mechanically: every `http(s)` URL in every merged page
-appears in a source PR's diff; every added body line in amended pages traces to
-a source PR hunk — orphan-line verification). Confidence fields were kept as the
-originating flushes set them, except client-side-rate-limiting where the union
-of provider-doc citations (Okta, Auth0, GitHub, OpenAI, RFC 6585) supports
-`verified` for the load-bearing claims. One subagent's fabricated content (12
-files matching neither main nor any PR, with invented source URLs) was detected
-by the same verification and replaced with true PR content.
+### A. `05d9b2c2` — the remediation diff is itself an unreviewed defect surface (qa)
+
+**Claim.** When a review-fix diff is handed to the next round, the edit's own
+shape opens defect classes the review never named: a widened fetch opens
+unbounded results, an added output line opens an unguarded path, a newly
+persisted record opens an orphan, a literal replaced by a placeholder opens
+unvalidated assembly.
+
+**Sources checked.** The general claim ("a fix is more defect-prone than the code
+it replaces") is already sourced on the target page to Yin et al., *How do fixes
+become bugs?*, ESEC/FSE 2011 — https://dl.acm.org/doi/10.1145/2025113.2025121 and
+https://www.eecg.utoronto.ca/~yuan/papers/incorrect_fix_abstract.html ("at least
+14.8% to 24.4% of sampled fixes for post-release bugs in these large OSes are
+incorrect"). The four-shape taxonomy is **field evidence**, not literature: PR
+#327 bot review rounds 13→14, where 4 of round 14's 7 warnings were defects
+introduced by round 13's own remediation, one per shape.
+
+**Confidence:** `verified` for the underlying claim (published study, quoted from
+the abstract page I opened); the four-shape table is recorded on the page as a
+dated **field observation**, distinguished from the cited study.
+
+### B/C. `269b1dad` + `df361a8b` — the receiving side of a completion gate (infrastructure)
+
+These two rows are the same insight harvested twice on 2026-08-12 (same trigger,
+same directive, different wording). Treated as one.
+
+**Claim.** When a Stop/completion gate repeats "verification loop incomplete" at a
+pause the worker's own prompt instructed, the worker holds its phase and reports
+the gate-vs-prompt mismatch; it does not advance the phase to a terminal value to
+silence the gate.
+
+**Verified by reproduction in this checkout at `fa89dc2`** (not from the
+installed plugin copy the candidates cited):
+
+| Check | Result |
+|-------|--------|
+| `hooks/loop-gate.sh:55` terminal set | `done\|approved\|merged\|failed\|""` — `impl_done` and `plan_ready` absent |
+| `skills/orchestrate/scripts/status-update.sh:6` phase vocabulary | `pending\|planning\|plan_ready\|implementing\|impl_done\|approved\|merged\|done\|failed` — `impl_done` is first-class |
+| `skills/orchestrate/templates/session-prompt.md:75,79` | instructs the worker to record `impl_done` "and wait" / "END YOUR TURN" |
+| `skills/orchestrate/scripts/ready-set.sh:74` | `satisfied()` = `approved\|merged\|done` — exactly the three values that would silence the gate |
+| `skills/orchestrate/scripts/ready-set.sh:28-30` | states the rule a fabricated phase breaks: "A dependency counts as satisfied only at `approved` or higher, NOT at impl_done" |
+
+So the harm is concrete and checkable, not rhetorical: the only phases that stop
+the nudge are the ones the scheduler reads as "reviewed", and writing one
+dispatches dependents against an unreviewed interface.
+
+**External source for the principle** — NIST SP 800-53 r5 AC-5, opened and quoted
+verbatim from https://csf.tools/reference/nist-sp-800-53/r5/ac/ac-5/ :
+"Separation of duties addresses the potential for abuse of authorized privileges
+and helps to reduce the risk of malevolent activity without collusion. Separation
+of duties includes dividing mission or business functions and support functions
+among different individuals or roles". The phase recording a review verdict
+belongs to the reviewing role; the reviewed party writing it is the same actor
+initiating and approving.
+
+**Confidence:** `verified` — reproducible in-repo at a named commit, plus a cited
+standard for the principle. The page's existing `confidence: verified` is
+unchanged; `last_verified` moved to 2026-08-13.
 
 ## Existing-layer check
 
-- Merged-main near-dup scan before consolidation: pairwise Jaccard over
-  title + "When this applies" across all 141 merged pages → **0 flagged pairs**;
-  previously merged content carries no duplication.
-- Cross-PR dedup during consolidation: 10 duplicate clusters collapsed to one
-  canonical page each (rate limiting 8→1, call-site enumeration 7→folded into
-  the canonical merged in #20, stderr/exit-0 diagnostics 4→1, sysroot 2→1,
-  env-off-switch 2→1, completion predicates 2→1, robots.txt 2→1,
-  harness-mediated results 2→1, leaked artifacts 2→1, orchestration category
-  naming unified). Three near-pairs kept distinct after trigger comparison,
-  with mutual `related:` links (differential setup vs interpretation; expansion
-  semantics vs off-switch design; import-time tactics vs level choice).
-- 24 existing pages received union-merged amendments; additions already present
-  in main (from #16/#20) were skipped, and all non-canonical `related:` ids
-  were remapped to canonical page ids (post-merge broken-link scan: 0).
+Routed via `INDEX.md` → `wiki/infrastructure/index.md` (agent-orchestration) and
+`wiki/qa/index.md` (process). Read every page in both categories whose "load when"
+overlapped, plus the qa page the fold targets.
+
+Pages read: infrastructure-agent-orchestration-session-completion-gates, infrastructure-agent-orchestration-control-signals-vs-primary-artifacts, infrastructure-agent-orchestration-shared-run-state, infrastructure-agent-orchestration-worktree-isolated-workers, infrastructure-agent-orchestration-pane-delivery-confirmation, qa-process-regression-scope, qa-process-release-gates
+
+**B/C — overlap found, merged rather than created.**
+`session-completion-gates` already covers this situation from the **gate author's**
+side: its own table classifies "instructed pause awaiting an external actor
+(`plan_ready` awaiting approval, `impl_done` awaiting review)" as terminal, and it
+cites the same `loop-gate.sh:55` line from a 2026-08-05 reproduction at `95cf947`.
+Its "When this applies" already says "when such a gate fires on a worker that did
+exactly what its own prompt told it to do".
+
+What it did **not** carry is the receiving side — what that worker should do while
+the gate is firing on it, and which phase values its role may write. That is the
+delta merged in:
+- "When this applies" extended to the worker's vantage point.
+- New **Do this** step 6 with a phase-authorship table (worker: `plan_ready`,
+  `impl_done`, and `done` only after approval; coordinator only: `approved`,
+  `merged`).
+- Two **Edge cases** rows (repeating nudge = mismatch not skipped work; cite both
+  file:line in the report) and one **Instead of** row (advance the phase → hold it
+  and report).
+- Two sources added: NIST AC-5, and the `fa89dc2` reproduction including the
+  `ready-set.sh:74` consequence the 2026-08-05 entry did not have.
+
+**No conflict.** The addition is consistent with the page's existing direction —
+the page tells gate authors to make instructed pauses terminal; the new step tells
+the worker not to route around the gap while it exists. Body 73 → 82 lines
+(limit 120). No banned qualifiers.
+
+**Related links.** No new `related:` entries: the four adjacent
+agent-orchestration pages were read and none is the counterpart of this delta
+(`control-signals-vs-primary-artifacts` is the orchestrator reading a worker's
+signals; `shared-run-state` is the file layout; `worktree-isolated-workers` is the
+brief contract; `pane-delivery-confirmation` is input delivery). The existing
+three-way `related:` set is already correct.
+
+**A — no page on main covers it; the fold target lives only in an open PR** (see
+below). `qa-process-regression-scope` is the nearest merged page and is a
+different axis (what to re-test for a *release*, by blast ring, not what a
+*remediation edit* opens). It was left untouched.
+
+## Open-PR check
+
+Listed 22 open `knowledge/*` heads (`gh pr list --search "head:knowledge/"`):
+#80 #79 #78 #76 #74 #73 #72 #69 #68 #66 #64 #62 #61 #58 #57 #56 #55 #52 #51 #50
+#49 #47. Pulled every one's diff (`gh pr diff N`) and grepped the added lines for
+each candidate's concepts.
+
+| Candidate | Overlapping open PR | Verdict |
+|-----------|--------------------|---------|
+| A — remediation diff as defect surface | **#76**, which adds `wiki/qa/process/defect-class-resweep-after-review.md`; its step 5 is "Read the remediation diff as unreviewed code", and its "Instead of" table already rebuts "treat the previous round's approval as covering the fix" | **fold** |
+| B/C — worker side of a completion gate | none. Only #79 touches `session-completion-gates.md`, and only to append one id to `related:`; #47/#51/#64/#80 merely name the page id in their reports or index rows. A grep of all 22 diffs' added lines for `impl_done`, self-approval, phase-fabrication and "advance the phase" returned no match (the only `forge` hits are #51's git-forge email page) | **new** (merged into the existing page here) |
+| C alone | duplicate of B — same trigger, same directive, harvested twice on 2026-08-12 | **drop** |
+
+**Fold executed, not deferred.** Pushed commit `4e9b0a1` to
+`dch0202-rsquare:knowledge/dch0202-rsquare-20260812-100014` (PR #76's head),
+adding a step 6 to that page — the four remediation shapes and the check each
+calls for — plus the PR #327 field-observation source. Explained on the PR:
+choiyounggi/dev-loop#76 (issuecomment-5274743196). That page stays at 80 body
+lines. **No sibling page was created here for candidate A.**
 
 ## Routing decision
 
-- New categories: `infrastructure/agent-orchestration` (5 pages; unified the
-  competing `orchestration`/`agent-orchestration` names), `databases/data-survey`
-  (1), `qa/deliverables` (1). All other pages route into existing categories.
-- Canonical-path decisions: rate limiting → `backend/common/reliability/`
-  (sits beside timeouts-and-retries; 6 of 8 variants chose it); stderr
-  diagnostics → `platforms/processes/` (concern spans beyond shells); leaked
-  artifacts → `testing/data/artifact-leakage-from-a-suite`; call-site
-  enumeration → the existing `backend/common/change-impact/` page.
-- All 38 new pages listed in their domain indexes (nearest-index rule; backend
-  routes via its python sub-index for bytecode-cache-staleness); INDEX.md domain
-  summaries updated for infrastructure/qa/databases. Full-wiki lint: frontmatter,
-  ids, related-links, index coverage, size, qualifiers, staleness → 0 findings.
+| # | Insight | Target | New category? |
+|---|---------|--------|---------------|
+| B/C | Worker side of a completion gate | `infrastructure/agent-orchestration/session-completion-gates.md` (**merged into existing page**) + `wiki/infrastructure/index.md` load-when extended + `log.md` | no — `agent-orchestration` owns worker/orchestrator protocol, and this page owns this exact situation from the other side |
+| A | Defect classes a remediation's own shape opens | `qa/process/defect-class-resweep-after-review.md` **on PR #76's branch** | no — routed to an in-flight page rather than duplicated |
+| C | (duplicate of B) | none — retired | — |
+
+Files changed in this PR: `wiki/infrastructure/agent-orchestration/session-completion-gates.md`,
+`wiki/infrastructure/index.md`, `log.md`, this report. No new page, no new category.
+
+Cross-Check: every URL cited here was opened in this session (the NIST AC-5 quote
+was fetched and quoted verbatim; the two Yin et al. URLs are pre-existing sources
+on PR #76's page and are quoted only as that page already quotes them). The
+in-repo file:line claims were re-run against this checkout at `fa89dc2` rather
+than inherited from the candidates' text, which cited the installed plugin copy.
@@ -43,3 +43,4 @@ Append-only. Format: `## [YYYY-MM-DD] <ingest|revise|lint|gap|contradiction|drif
 ## [2026-08-06] revise | Union-merged amendments from the same 15 PRs into 24 existing pages (largest: tests-that-cannot-fail — 7 PRs: per-assertion mutation granularity, restore-mechanism-by-commit-state, suite-total-drop detection; portable-shell-scripts — 8 PRs: POSIX inline set-- reordering, ${VAR:-} vs ${VAR-} colon semantics, quote-by-what-the-text-is; test-data-and-isolation — 6 PRs: env-derived write paths, absent-variable cases, exec-bit fixtures, leak attribution; change-impact/call-site-enumeration — 8 unique additions from 7 path-variants: rename-in-place, same-type reorder, *args forwarding, two-searches-agree fallacy, codemod, coverage completion check). All non-canonical cross-page ids remapped to canonical pages; every added line traced to its source PR hunk (fabricated worker output was detected by orphan-line verification and replaced with true PR content).
 ## [2026-08-06] dedup | Cross-PR duplicate clusters collapsed: client-side rate limiting (8 PR versions at 3 paths → 1), call-site enumeration (7 versions at 7 paths → additions folded into the already-merged canonical page), stderr/exit-0 diagnostics (4 versions at 4 paths → 1), macOS sysroot (2→1), env-var off switch (2→1), completion predicates (2→1), robots.txt (2→1), harness-mediated results (2→1), leaked test artifacts (2→1), orchestration category naming (orchestration vs agent-orchestration → agent-orchestration). Kept distinct after trigger comparison: differential-testing vs differential-run-agreement (setup vs interpretation), unset-versus-empty-parameters vs env-var-off-switches (semantics vs design, cross-linked), import-time-side-effects vs test-level-choice edge (tactics vs level choice, cross-linked). Merged-main near-dup scan (Jaccard over title+trigger, 141 pages): 0 pairs — no duplication among previously merged content.
 ## [2026-08-06] ingest | Reconciled post-consolidation flush PRs #42–#43 (created against pre-#44 main). Folded unique content: platforms/filesystems/permissions-and-exec-bits (+`sh "$SCRIPT"` interpreter-invocation rows — stubs need no exec bit, avoids EDR chmod+x heuristics), platforms/shells/command-text-inspected-before-execution (+gate-AUTHOR step 8: parse all three POSIX quoting forms, expand only ~/$HOME/${HOME}; red-then-green bats evidence), testing/quality/checks-that-cannot-pass (+vacuously-green pre-implementation test row), platforms/processes/tool-diagnostics-without-a-failing-exit-code (+repeat-suppression edge, +clang reproduction source), infrastructure/ci-cd/secrets-handling (+independent credential channels row: gh token vs git SSH vs API token, git push --dry-run / author:@me probes), testing/mocking/what-to-mock (+ESM read-only-bindings DI rows, +PATH-emptied tripwire negative-proof). Dropped as duplicates of the #44 state: #42's rate-limit row (already in timeouts-and-retries + client-side-rate-limiting), #42's warning-only-diagnostics page (same trigger as tool-diagnostics-without-a-failing-exit-code), #42's leak-attribution row (already in test-data-and-isolation).
+## [2026-08-13] ingest | knowledge-flush of 3 queued insights: 1 merged into an existing page, 2 retired. Merged into infrastructure/agent-orchestration/session-completion-gates: the receiving side of a completion gate — a worker on which the gate repeats at an instructed pause holds its phase and reports the gate-vs-prompt mismatch rather than advancing to a terminal value, with a phase-authorship table (worker writes `plan_ready`/`impl_done`, and `done` only after approval; `approved`/`merged` are the coordinator's review verdict). Reproduced at `fa89dc2`: `hooks/loop-gate.sh:55` accepts `done|approved|merged|failed|""` while `skills/orchestrate/scripts/status-update.sh:6` lists `impl_done`, and the three values that would silence the gate are exactly the three `skills/orchestrate/scripts/ready-set.sh:74` counts as dependency-satisfying — so a fabricated phase dispatches dependents against an unreviewed interface. Sourced to NIST SP 800-53 r5 AC-5 (separation of duties). Retired: the second copy of that insight (same trigger, same directive, harvested twice in one day) and the review-remediation defect-surface insight, folded into open PR #76's qa/process/defect-class-resweep-after-review rather than re-ingested here.