# Row 70 round 2 (#1564): docs for `mosaic` on PATH Sage, 2026-10-15. Answers Filbert's round 1 verdict, comment 28598 (rev 1326). Same brief, still pinned at blob d56345b76ffd (rev 1314). - Base: 48b893871 (`base.txt`), as in round 1. No commit since then touches a manifest path (`git log 48b893871..HEAD -- <4 paths>` is empty at 26f4f9701). The landing gate runs on HEAD. - Candidate: `build.patch` (whole change from base), `candidate-manifest.sha256`, the same 4 files as round 1. Manifest file sha256 `098b9aefe830c53173f0b673705e2fddfa5b49a9ba23903a045bf72d8875604e`. - `delta.patch`: round 1 tree to round 2 tree, file by file. ## What changed since round 1 | Finding | Change | |---------|--------| | F1 (blocking): the `scripts/mosaic` header doesn't name `mosaic-old` | One comment sentence added: "On Jason's host the estate CLI it replaced is `mosaic-old`." No code line moved or changed. | | Doc-only 1: TOOLS.md names neither `mosaic-old` nor the CLI README link | The seat-launch paragraph now names `mosaic-old` and links `../packages/cli/README.md#one-pinned-copy-row-85-1579`, the anchor TOOLS.md already uses at line 357. The exit-code sentences after it are only rewrapped. | | Doc-only 2: the seat README omits TOOLS.md "Stack copies" | The wrapper sentence now links `../../docs/TOOLS.md#stack-copies-scriptsstack-copysh`. | | Non-blocking: the test comment says "Comment lines" | Now "Lines that start with "# " are left out", and it says the shebang and any other line stay in. | The two follow-ups were cheap and in files this row owns, so I did them in this round rather than leaving them open. ## Checks - `bash -n scripts/mosaic` passes. `git diff --check` is clean. No added line has an em dash or curly quote. Two added lines exceed 80 columns; both are a single Markdown link that can't wrap. - Both new link targets exist: `docs/TOOLS.md:279` "## Stack copies (`scripts/stack-copy.sh`)" and `packages/cli/README.md:466` "### One pinned copy (row 85, #1579)". - `node --test packages/queue/tests/dispatch.test.mjs` in the round 2 tree: pass 4, fail 0. - Mutants on round 2's `scripts/mosaic`, each run against the pre-A2 test: `set -euo pipefail` commented out as `# set -euo pipefail`, `REPO=` dropped, shebang dropped. Each fails it (fail 1). Control: dropping header line 12 passes (fail 0), which is the narrowing working as intended. The file was restored and `cmp` matched. - No full gate for this round. Round 1's second gate passed all 24 suites on 48b893871 (`../gate-run.log`). Round 2 changes comments and Markdown only, and the dispatch test that reads them passes. The landing gate reruns everything on HEAD before the commit.