Garry's Opinionated OpenClaw/Hermes Agent Brain
TypeScript
Merge remote-tracking branch 'origin/master' into capy/gbra-224-w5-a
garrytan5ddd1c865108
diff --cc CHANGELOG.mdindex cc4e9b76b0,13fc31a1c0..0886aae997--- a/CHANGELOG.md+++ b/CHANGELOG.md@@@ -10,25 -10,19 +10,38 @@@ credits are retained; no result has beeidentifiers and attribution are available in the pre-removal Git revision`6040075c6cb95be5881cc2e1b76ef7d71f4e5d29` (retained on 2026-09-23).+## [Unreleased]++**A fresh install and an upgrade add nothing you didn't turn on.**++The default install (keyless PGLite, no repo) already started no daemon and installed no service. That is now locked by tests for `gbrain setup claude-code`, `gbrain setup codex` and `gbrain init --pglite --no-embedding`: no process outlives the command, no systemd unit, launchd plist, crontab entry or autopilot script is written, no connector is connected, and self-upgrade stays in `notify` mode.++One upgrade path did install a daemon. On a Postgres brain with `sync.repo_path` set and no autopilot, the v0.11.0 migration's Phase F ran `gbrain autopilot --install`. That migration runs on `apply-migrations --yes`, `post-upgrade` and package postinstall, so a brain that never asked for autopilot could get one. Phase F now installs nothing on a brain without an autopilot: it is skipped (reason `not_installed`), prints the install command and records it, and `gbrain doctor` shows it as a `minions_migration` line until you run it. A brain that already runs autopilot gets it reinstalled as before, and an upgrade still rewrites only a service unit that is already installed.++### For operators++- INSTALL_FOR_AGENTS.md has one **Opt-ins** list (keys and embeddings, autopilot, dream schedule, connectors, auto-upgrade, release channel), one command each, and `gbrain setup <harness>` prints one line pointing to it.+- To get autopilot on a brain that has none, run the command doctor shows: `gbrain autopilot --install --yes --repo <path>`.+- `docs/integrations/reliability-repair.md` is linked from llms-full.txt instead of inlined, to stay under the size budget.++### Tests++- `test/boring-default.test.ts`: fresh setup (both harnesses) and init add nothing; `post-upgrade` without the opt-out on repo-less and repo brains installs no unit; an installed unit is the only one rewritten.+- `test/apply-migrations-safety.serial.test.ts`: Phase F with `sync.repo_path` set and no autopilot installs none (the postinstall argv); with one installed it reinstalls as before.++ ## [0.60.477.0] - 2026-10-28++ **A trust release confirms only what its preview showed.**++ Releasing a held fact that already sits in a page's `## Facts` table confirmed that row's confidence, notability, validity and other cells, though the owner's preview showed only the text (#7656). The database-only release, which every brain without a repository uses, could also release a hold that was re-gated after the owner approved it (#7657).++ ### What changed++ - **The preview names every field the release confirms.** That's the held row's cells as they stand, or the held payload's kind, visibility, validity, source and context. The approval binds them, so a row edited after the approval refuses `preview_changed` and the fact stays held.+ - **The preview can't be made to show a field the row doesn't have.** A held fact's text is printed quoted and escaped, any other value containing a quote, comma or semicolon is quoted the same way, and invisible or bidirectional formatting characters become spaces.+ - **The database-only release re-checks the approval** on the locked hold and refuses `preview_changed` when the hold changed after the owner approved it, the same as the managed path.+ - **New doctor check `trust_release_unpreviewed_fields`** lists facts released from a held row before this fix, so the owner can review them. It's read-only and never lowers a tier.+## [0.60.476.0] - 2026-10-27**Text someone else wrote can no longer add facts to your brain.**