[entelecheia]

Closes #94

Final T4 after #120, #121 and #122. Command/module text follows the supplied writer instead of bypassing it through process stdout. RunContext.Out resolves through Output() with the previous stdout default when nil. CLI contexts pass cmd.OutOrStdout(); console logs and apply-recording warnings use the error writer. Metadata/audit stderr warnings remain separate. Explicit tenant writers (including discard) retain priority.

Module and CLI bare stdout prints are gone. Shared UI helpers cover styled status rows and plain text/tables; names-only lists and raw id/groups/quota payloads keep their bytes. Production glyph literals are centralized in UI, including unstyled constants where a description enters JSON. Tenant uses the leaf UI constants; low-level backup previews use ASCII -> rather than adding an exec-to-UI dependency. Fleet shares the JSON encoder with its existing HTML-escaping mode and uses the same exit-code constant.

Verification

• AC1/AC2: preceding collector/module-name work is retained. All 35 T3 status/check/doctor byte goldens are unchanged; the full race suite passes them. A new encoder test pins native versus fleet HTML escaping.
• AC3: an AST/source guard verifies no bare fmt.Print* in module/CLI production code and no production glyph literal outside UI. Explicit-writer, nil fallback, stdout-leak, warning separation, raw-payload and table tests cover the routing.
• AC4 / contract chore(release): staged fleet rollout of the epic #98 release without disruption #115: remote exact-head full race/build/vet passes at 11455ae, with failing ambient ufw and sshd first on PATH. Twenty managed renderers/constants are byte-identical to main. Config/profiles/fingerprints, fleet remote commands, sudoers and schedule sources are untouched. Six-profile module tests assert actual Check/RunAll outcome order and unchanged managed directories/files/backups.
• Six-profile full CLI captures use isolated binaries with only existing detector/path seams: Ubuntu/Rocky SystemInfo, private SSH/Docker files and private command fixtures. Every selected module reaches RunAll; all twelve before/after invocations exit 0. Stdout differs only in marker spacing; stderr is identical after query-log clock normalization. File/state/lock/audit checks pass. These are controlled fixtures, not native/live fleet proof.
• Specs: 21, 0 errors, 1 existing goal warning. Release notes: 32 pass. Source/gofmt checks pass.
• Reviewer independently completed the previously deferred Mac check at 11455ae after the host became healthy: failing UFW first on PATH, race tests for CLI/module/UI/tenant/exec passed (14.5s/65.6s/1.4s/2.6s/2.2s), go vet ./internal/... clean, specs 21/0 errors. The author's earlier admission deferral is superseded by this exact-head reviewer evidence.

Reviewed text diffs

All six profiles have before/after captures under internal/module/testdata/output-t4/ and full CLI captures under internal/cli/testdata/output-t4/ (plus exit/stderr results). With colors disabled, the common change is one additional space after the status marker:

-  ✓ locale: already satisfied
+  ✓  locale: already satisfied
-  → system: Cap journald at 2G
+  →  system: Cap journald at 2G

When colors are enabled, RunAll uses the existing styled UI markers. Raw payload/table spacing and command words remain unchanged. Rollback symlink previews use ->; machine-facing Storage and Cloudflared arrows retain their existing bytes through unstyled constants. The fixture seeds a chosen Docker log driver so the pre-existing daemon-updates map iteration does not make captures nondeterministic; output is never sorted to hide module ordering.

Release notes

Text status rows use the common marker spacing/style, and rollback symlink previews use an ASCII separator. Commands embedded with a custom output writer now capture their module text consistently. JSON/Prometheus, managed file content and fleet/schedule commands are unchanged. No new reload, restart or operator action is introduced.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has insufficient remaining quota. Try requesting a review at a lower effort level.

[chatgpt-codex-connector]

Codex Review Summary: ✅ Completed

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-11-09T22:30:41.599756Z	11455ae	Draft marked ready

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.