choiyounggi · choiyounggi · Aug 6, 2026
@@ -1,53 +1,116 @@
-# Knowledge consolidation — 15 open PRs (#17–#40) → one reconciled state
+# Knowledge flush — 11 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.
+Drained 9 queue files (11 candidate rows) from `~/.dev-loop/queue/`. 5 candidates
+were exact duplicates of knowledge already ingested by the #42–#46 consolidation
+and were retired without edits; 6 were merged into existing pages (merge-before-create
+— no new pages, no new categories). Open-PR dedup check: `gh pr list --label
+dev-loop:knowledge --state open` returned zero open PRs, so nothing here overlaps
+an in-flight review.
 
 ## 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.
+**1. Bats assertions as `[[ ]]` mid-test are decoration on bash 3.2 → `verified`.**
+Claim: under bats on macOS system bash (3.2.57), a false `[[ ]]` that is not the
+test's last command does not fail the test; `[ ]` and `grep -qF` pipelines fail at
+any position. Verified this session by fresh local reproduction (Bats 1.14.0, GNU
+bash 3.2.57, arm64): 4-test probe file gave `ok` for mid-test false `[[ ]]`,
+`not ok` for mid-test `[ ]`, mid-test grep, and last-line `[[ ]]`. Mechanism
+isolated outside bats: `bash -ec '[[ … ]]; echo survived'` exits 0 while the `[ ]`
+form aborts — a pre-4.0 errexit semantic, not a bats defect. Sources:
+https://tiswww.case.edu/php/chet/bash/COMPAT (bash-4.0 changed `set -e` to exit on
+compound-command failure), https://bats-core.readthedocs.io/en/stable/gotchas.html
+and https://www.shellcheck.net/wiki/SC2314 (the documented same-shape gotcha for
+negated `!` commands: they "can never fail when used in the middle of a test").
+
+**2. Widened scan surface turning old tests red → triage as first true positive → `field-tested`.**
+Claim: when a leak/masking detector gains a previously unscanned output channel and
+an existing test reddens, diff what the new surface saw against what the fixture
+declares, and fix the fixture's smuggled data rather than the detector. Evidence is
+the session reproduction (linkly #43: `result.bindings` widening reddened
+`test_when_guard_removed_diverges`; fixture carried an undeclared `password` key;
+narrowing to declared fields → 1218 green). No independent external source claims
+this exact triage order, so it stays field-tested; it is consistent with the page's
+existing sourced principle (change-detector tests / "the guard reddens on a genuine
+S-and-C artifact means the guard is working").
+
+**3. Worker usage-limit pause looks alive → check pane tail for the limit marker → `field-tested`.**
+Claim: simultaneous quiet workers with green liveness are a usage-limit pause;
+find the `You've hit your session limit · resets HH:MM` marker, then after reset
+send a resume prompt ordering state-recheck → remaining DoD → completion signal.
+Evidence: 2026-08-06 run, three workers paused on one reset with identical markers;
+the structured resume prompt recovered all three at their exact interruption point.
+Vendor docs do not document the marker string, so no external citation is possible —
+kept field-tested with the context described on the page.
+
+**4. Dispatch after `worker_done` needs a substrate-idle wait; a failed dispatch consumes the task → `field-tested`.**
+Evidence: reproduced twice in the 2026-08-06 run (immediate dispatch →
+`runtime_unavailable` + task consumed; dispatch after `orca terminal wait --for
+tui-idle` succeeded first try). Orca is an internal tool; no external source exists.
+
+**5. Guardrail `worktree_escape` can escalate on read-only cross-worktree access → `field-tested`.**
+Evidence: two read-only commands (`awk`/`grep` over an upstream FINDINGS file,
+`git status`) each raised `ask` and stopped the watch with exit 5; both approved
+after review. This **conflicts** with the existing page's 1.0.0 reproduction where
+reads passed — handled as a condition-dependent (rule-version) edge case, not an
+overwrite (see below).
+
+**6. `sh "$SCRIPT"` stub seam under EDR — second reproduction → enriches existing `field-tested` row.**
+Directive already on the page from the #42–#43 reconciliation; this flush adds the
+independent second reproduction (8 stall-handler stubs injected without any
+`chmod`, bats suite 331/331 green under SentinelOne).
+
+**Dropped as exact duplicates (no edit, retired from queue):** gate quoting-form
+parsing (already `command-text-inspected-before-execution` step 8 + Instead-of row
++ field context citing the same bats tests 12–13), stderr-warnings-with-exit-0
+capture incl. redirection order (already the whole of
+`tool-diagnostics-without-a-failing-exit-code`), Homebrew clang `-isysroot
+$(xcrun --show-sdk-path)` (already the whole of `compiler-sysroot-on-macos`, incl.
+the same 69-failure repro), temp-artifact prefix counting + AST-rule enforcement +
+RED-first guard (already `artifact-leakage-from-a-suite` steps 1/4/5), and
+vacuously-green pre-implementation usage-error test proven by guard mutation
+(already a `checks-that-cannot-pass` edge row describing the identical
+unknown-subcommand/exit-1 case).
 
 ## 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).
+Read before deciding: root `INDEX.md`; domain indexes for testing, platforms,
+infrastructure; and the nine candidate-overlapping pages
+(`command-text-inspected-before-execution`,
+`tool-diagnostics-without-a-failing-exit-code`, `compiler-sysroot-on-macos`,
+`artifact-leakage-from-a-suite`, `worktree-isolated-workers`,
+`tests-that-cannot-fail`, `checks-that-cannot-pass`, `guard-shape-vs-consequence`,
+`control-signals-vs-primary-artifacts`, `permissions-and-exec-bits`,
+`destructive-operations-on-shared-daemons`).
+
+- **Merged, not created:** all 6 surviving insights landed as edge-case/table rows
+  and evidence on existing pages. No new page, no new category.
+- **Conflict flagged and resolved as condition-dependent:**
+  `worktree-isolated-workers` states reads pass the guardrail (1.0.0 repro); the
+  new observation shows a rule version escalating on reads. Added as an edge row
+  ("guardrail rules differ by version — probe one read before fanning out") and
+  noted in the log entry; the Do-this table was not overwritten.
+- **Duplicates:** the 5 dropped candidates matched existing pages
+  trigger-for-trigger and directive-for-directive (the #42–#46 reconciliation had
+  already ingested earlier harvests of the same sessions' insights).
+- **Related-links:** no new cross-links needed — every edited page already links
+  the pages the new rows reference (e.g. `tests-that-cannot-fail` ↔
+  `checks-that-cannot-pass`, `control-signals` ↔ `worktree-isolated-workers` via
+  the agent-orchestration index).
 
 ## 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 (existing page) | Why this page |
+|---------|------------------------|---------------|
+| bats/bash-3.2 mid-test `[[ ]]` | `testing/quality/tests-that-cannot-fail` — new never-fails-pattern row + sources | The page owns "assertions that cannot detect a defect"; this is a shell-level instance of that exact class |
+| widened-scan-surface red | `testing/quality/guard-shape-vs-consequence` — When-this-applies clause + edge row + field evidence | The page owns guard-red triage; existing edge row already covered "guard reddens on genuine S-and-C"; this adds the widened-surface trigger and fixture-diff triage |
+| usage-limit worker stall | `infrastructure/agent-orchestration/control-signals-vs-primary-artifacts` — edge row + field evidence | The page owns done/alive/stalled/dead verdicts; this is a new stalled-state cause with recovery protocol |
+| dispatch-after-done timing | same page — edge row | "Done signal ≠ substrate release" is precisely the page's signal-vs-artifact distinction |
+| read-only guardrail escalation | `infrastructure/agent-orchestration/worktree-isolated-workers` — edge row (condition-dependent conflict) | The page owns the guardrail's read/write asymmetry; the conflicting observation must sit next to the claim it qualifies |
+| EDR stub second repro | `platforms/filesystems/permissions-and-exec-bits` — evidence sentence | Directive already lives there; only evidence strengthened |
+
+Domain hints from the queue were respected except where a page already owned the
+case: the "testing"-hinted EDR-stub insight routes to platforms (the page that owns
+exec-bit/EDR invocation style), and the "platforms"-hinted orchestration insights
+route to infrastructure/agent-orchestration (dedicated category), consistent with
+prior flushes. Index "load when" lines updated for the four pages whose routing
+surface grew.
@@ -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).
+## [2026-08-06] ingest | knowledge-flush of 11 queued insights (9 queue files). Dropped 5 as exact duplicates already folded by the #42–#46 state: gate quoting-form parsing (command-text-inspected-before-execution step 8), stderr-with-exit-0 diagnostics (tool-diagnostics-without-a-failing-exit-code), macOS keg-only clang sysroot (compiler-sysroot-on-macos), artifact-prefix leak attribution + AST-rule enforcement (artifact-leakage-from-a-suite), vacuously-green pre-implementation test + mutation proof (checks-that-cannot-pass edge row). Merged 6 into existing pages, no new pages: testing/quality/tests-that-cannot-fail (+bash-3.2 mid-test `[[ ]]` never-fails row — locally reproduced this session on Bats 1.14.0/bash 3.2.57 and mechanism-isolated to pre-4.0 errexit via bash COMPAT; verified), testing/quality/guard-shape-vs-consequence (+widened-scan-surface first-true-positive edge, linkly #43 result.bindings fixture repro; field-tested), infrastructure/agent-orchestration/control-signals-vs-primary-artifacts (+usage-limit-pause stall edge with resume-prompt protocol, +dispatch-after-done tui-idle/consumed-task edge; field-tested 2026-08-06 Orca run), infrastructure/agent-orchestration/worktree-isolated-workers (+read-only-escalation guardrail-variant edge — condition-dependent conflict with the 1.0.0 read-pass reproduction, resolved as an edge row per version-dependence rather than overwriting the Do-this table; field-tested), platforms/filesystems/permissions-and-exec-bits (+second `sh "$SCRIPT"` stub reproduction, bats 331/331 without chmod). Index load-when lines updated for the four edited routed pages.
@@ -8,7 +8,7 @@ sources:
   - https://man.openbsd.org/tmux
   - https://code.claude.com/docs/en/hooks
   - https://pubs.opengroup.org/onlinepubs/9699919799/utilities/V3_chap02.html
-last_verified: 2026-08-05
+last_verified: 2026-08-06
 related: [infrastructure-agent-orchestration-shared-run-state, infrastructure-agent-orchestration-pane-delivery-confirmation, infrastructure-agent-orchestration-session-completion-gates, platforms-shells-command-text-inspected-before-execution, platforms-tools-harness-mediated-tool-results, debugging-methodology-hypothesis-testing]
 ---
 
@@ -67,6 +67,8 @@ progress by running a script that writes outside its worktree.
 | An editor/Write tool succeeds where Bash was refused for the same path | The guardrail inspects Bash command text only, so the two channels disagree by design. Do not use the working channel to route around the rule — report it |
 | Heartbeat is fresh but no commits in N intervals | Fresh heartbeat plus unchanged primary artifact is the stalled state, distinct from alive-and-progressing and from dead; handle it as its own case |
 | The monitor is the only thing that can see the worker | Add the primary-artifact check to the monitor rather than trusting its verdict; a monitor with no artifact check cannot produce evidence |
+| Several workers go quiet almost simultaneously: liveness checks pass, heartbeats stay fresh, diffs stop growing | Before restarting anything, search each worker's pane/terminal tail for the CLI's usage-limit marker (e.g. `You've hit your session limit · resets HH:MM`) — a usage-limit pause is idle waiting, not a crash, so every liveness probe passes. After the stated reset time, send a resume prompt that orders: re-verify state (`git status`, rerun the tests) → remaining definition-of-done → completion signal. A bare "continue" sent before the reset is consumed by the same limit message, and a resume without the state re-check makes the worker guess where it stopped |
+| The next task is dispatched to the same terminal immediately after the worker's done signal and fails runtime-unavailable | The done message is the worker's report time, not the substrate's release time — the CLI is still tearing down its stop-hook chain and the previous dispatch still occupies the terminal. Wait for the substrate's own idle signal (e.g. `orca terminal wait --for tui-idle`) before dispatching; and when the failed dispatch consumed the task, create a new task from the same spec — the consumed one cannot be retried |
 
 ## Instead of
 
@@ -82,3 +84,4 @@ progress by running a script that writes outside its worktree.
 - https://man.openbsd.org/tmux — `has-session`, `display-message` behavior and session naming
 - https://code.claude.com/docs/en/hooks — hook architecture and path guardrails
 - https://pubs.opengroup.org/onlinepubs/9699919799/utilities/V3_chap02.html — exit codes and output redirection
+- Field evidence 2026-08-06 (Claude CLI workers under tmux/Orca orchestration): three workers paused simultaneously on one usage-limit reset — identical `You've hit your session limit · resets 01:10` marker in each terminal tail while every liveness check passed; a state-recheck resume prompt sent after the reset resumed all three exactly at their interrupted step (test re-run). Same run: two dispatches issued immediately after `worker_done` both failed `runtime_unavailable` and consumed their tasks; dispatches issued after a `tui-idle` wait succeeded first try