loki-mode 10.6.13 → 10.6.14

package/docs/v10/DECISIONS.md

@@ -588,3 +588,14 @@ The $20.00 default per-run cap stays. It is a hard stop on priced sessions only;
 
 ## D79
 2026-10-03T10:48Z, CoS (D69). S-218r round 5: the untracked-status snapshot needs git 2.44 or newer. On an older or unparseable version it refuses, returns 1 and prints a clear message; the callers already treat a failed snapshot as no snapshot. Why: on older git the repo's own config can fake the GIT_CONFIG_COUNT sentinel, and GIT_NO_LAZY_FETCH does not exist, so a hostile repo could run a filter driver or an ssh command (both reproduced by opus). Apple git on current macOS is 2.54; Ubuntu 22.04 (2.34) and Debian 12 (2.39) degrade to no untracked snapshot rather than an unprotected one.
+
+## D80
+2026-10-03T12:54Z, CoS (D69), moat change pending CTO ratification in the SEAL r9 opus review. SEAL r8 BLOCK: in a single-process run (`node --test --test-isolation=none`, or a test file run directly with `node file.js`) a test can write a complete fake summary block and call process.exit(0) before the runner prints its own, so a failing test that never ran is reported VERIFIED. No output parsing can tell the difference, because the test wrote the whole output. Decision: (1) loki-seal grants test coverage only when the npm test script is `node --test` with process isolation (the default), where the parent runner always prints its own summary after the child exits; isolation=none, a direct `node file.js` script, or any script it cannot classify as that gives NOT VERIFIED with a stated reason. (2) As a second layer, a test file that calls process.exit, process.reallyExit, process.abort or process.kill is an integrity finding (NOT VERIFIED). (3) The truncation check gets its own unit test that does not rely on the end-of-output rule. Over-refusing honest single-process setups is accepted (fail closed).
+
+## D80 amendment (2026-10-03T13:14Z, CoS per D69, from the SEAL r9 opus review)
+The r9 reviewer amended D80 and the CoS adopts the amendment as binding. The reason is a reproduced finding: the classifier matched `node\s+--test` anywhere in the script, so "node --test-reporter=spec t.js", "node --test-only t.js", "node --test && node fake.js" and "echo node --test; node t.js" were all classified as node --test (a node -e run at 13:13Z). Amended rule:
+- loki-seal grants coverage from npm test only when the whole trimmed script is exactly `node --test` plus allowlisted flags (--test-reporter=<name>, --test-reporter-destination=stdout, --test-isolation=process, --test-concurrency=<n>, --test-timeout=<n>) and plain path arguments.
+- NOT VERIFIED, with a stated reason, applies on: any other token or a shell metacharacter (; & | $ ` quotes # < >); a pretest or posttest script; a project .npmrc setting node-options or script-shell; or NODE_OPTIONS set in the hook environment.
+- The exit/abort/kill scan remains as defense in depth and is not the trust boundary.
+- The truncation check keeps its own unit test.
+The CTO ratifies this in the r10 HIGH review.

package/docs/v10/PROGRESS.md

@@ -2055,3 +2055,39 @@ The swarm is stopped: .loki/V10-STOP exists, the loop has no wake-ups, and the c
 - Merged: P7-SCALAR-PIN d7711bb0e (opus APPROVE, 13 mutations, new fixtures catch 11; P7 moat script rc 0 on main).
 - Opus BLOCKs: SEAL-FORGED-LINES r4 (TODO-shaped name, newline name, describe.todo and failing describe each free a forged slot on spec). D69 ruling: spec path fails closed (NOT VERIFIED) on any ambiguity, TAP named as the sound reporter; r5 building. W1-S2 r4 (dotted stems, path vs basename tie) and r5 (index/__init__ dir modules, suffix paths ranked exact). D69 ruling: bias to over-exclusion. r6 e1083d661 in opus review.
 - Building: SEAL r5, D61-11 r3. P0-CORE-BUDGET TL r2 still open.
+
+## 2026-10-03T11:57Z CoS
+- v10.6.12 on npm latest at 11:53:55Z (registry checked); peer notified with the user-facing change list.
+- Train/97 pushed at 80665d4ca (P7-SCALAR-PIN d7711bb0e plus docs); CI waiting on Tests, Bun Parity, Coverage, Security Audit.
+- Opus BLOCK: SEAL-FORGED-LINES r5 f02531df5, B1 reproduced: a test name with \r, U+2028 or U+2029 is unparsed by SPEC_MARK and silently dropped, freeing a forged pass slot (PASS on default and spec; TAP holds). D69 ruling: any mark-prefixed line that does not fully parse is ambiguous, NOT VERIFIED. r6 building (sonnet).
+- W1-S2 r8 05be797c5 in opus review. D61-11 r3 builder over budget (451 min), finishing its mutation check; re-slice if it misses this cycle. P0-CORE-BUDGET TL r2 still open.
+
+## 2026-10-03T12:20Z CoS
+- v10.6.13 (train/97, af05ca299) released by train-cycle at 12:08Z; release.yml still in_progress at 12:18Z, npm latest still 10.6.12.
+- Merged on local main since: PO3 HELP-ALLCMDS, HEAL-DOC, PAUSED-TTY, MEM-INDEX, SKILL-RUNTIME-DOC, COUNCIL-TODO; next train cuts after npm promotes 10.6.13.
+- PO4 refill landed 9 cards (BOARD a57e57635); 7 sonnet builders dispatched; CLI-SUITES-REG waits on START-REPO-DIR-FIX, TODO-PLURAL (HIGH) waits on COUNCIL-TEST-TMP.
+- W1-S2 r11 opus BLOCK (unicode names in plain import list leak; r10 same); r12 fix building. SEAL r7 11019f987 113/0, in opus re-review.
+
+## 2026-10-03T12:50Z CoS
+- train/98 reds were two HELP-row test pins (cli.ts:47); fixed by 6c894fce1 and 2e1d79095, no slice dropped.
+- Ten slices merged (PO3-QS-MATCH, nine PO5); train/99 pushed at 12:47Z (2e6613aac), CI pending.
+- PO5-HELP-READONLY merged as 4eba90e3f for train/100.
+- P0-CORE-BUDGET TL r2 BLOCK not reproduced: test-engine10-dist.sh 18/0 on main.
+- Opus HIGH re-reviews running: W1-S2 r15 (f9dc7d67b), D61-11 r2 (535552d74). SEAL r8 builder running. W1-S3 waits on W1-S2 (shares wall.ts).
+
+## 2026-10-03T13:09Z CoS
+- P0-SIZE-BUDGET merged as 87f256642. budget.test.ts passes 27/0 and test-release-dist-guard passes 13/0, so main is back under its size caps.
+- W1-S3: the builder fixed TS2345 in 1e630cc08 (sizing.test 22/0). A TL MEDIUM review (sonnet) is running.
+- D61-11b: r2 fix for the opus BLOCK has been requested.
+- SEAL: r9 per D80 is still building.
+- train/99 (2e6613aac): Tier A, Bun Parity, Security, First-run and Coverage are green. Tests failed on shards 3, 4 and 7, all in issue/start routing suites: test-issue-to-pr, test_start_run_unified, and test-start-repo-directory (PO4).
+  - Locally on main, all three pass with rc=0.
+  - No train/99 commit touches the issue parser or provider detection.
+  - Full rerun started at 13:08Z per D68. If it fails again, open a P0.
+
+## 2026-10-03T13:18Z CoS
+- D61-11b merged (f6dea72bd, 5e6b5e9c3, dist 095f17e79) after an opus r2 APPROVE. On main: budget, unit_mode, features and supervisor 320/0; tsc 0; dist guard 13/0.
+- SEAL r9 BLOCK: the reviewer's desk finding reproduced by node -e. The classifier matches `node --test` as a substring. D80 amended (c7d9d02df). r10 requested.
+- G-04 built (0a90538a8, 25/0). Opus HIGH review is running.
+- W1-S3: TL BLOCK, because its wiring test survives a mutation. A test-only r2 fix was requested.
+- train/99 Tests: the full rerun failed on the same shards 3, 4 and 7, so the failure is real. P0-T99-ROUTING is open and building (sonnet). train/99's release and train/100 are held until it is green.