chore(deps): bump the root-npm group across 1 directory with 3 updates - #137
Open
dependabot[bot] wants to merge 3 commits into
Open
chore(deps): bump the root-npm group across 1 directory with 3 updates#137dependabot[bot] wants to merge 3 commits into
dependabot[bot] wants to merge 3 commits into
Conversation
dependabot
Bot
added
dependencies
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
dependabot
Bot
changed the title
dependabot
Bot
force-pushed
the
dependabot/npm_and_yarn/root-npm-ce52c755a9
branch
from
August 10, 2026 16:25
6105602 to
b378847
Compare
dependabot
Bot
force-pushed
the
dependabot/npm_and_yarn/root-npm-ce52c755a9
branch
from
August 10, 2026 17:01
b378847 to
fc54b61
Compare
stuinfla
added a commit
that referenced
this pull request
Aug 12, 2026
…configured is not operational
ground-ruvnet.sh treated ANY occurrence of RUFLO_HARNESS_LOOP in .claude/settings.json as
proof the self-learning flywheel was on, and then went silent. Verified against rUv's
source at f35c545 rather than recalled:
harness-worker.ts:41 /^(1|true|yes|on)$/i.test(process.env.RUFLO_HARNESS_LOOP ?? '')
harness-worker.ts:54 if (!optedIn) -> reason: 'opt-in required (RUFLO_HARNESS_LOOP=1)'
harness-project-anchor.ts:186 'project-local flywheel anchor required; create …'
harness-project-anchor.ts:32 DEFAULT_PROJECT_ANCHOR_MANIFEST = .claude/eval/flywheel-anchor.manifest.json
The daemon reads its OWN process environment. A Claude settings file does not put anything
into the environment of a daemon launched by Codex, launchd, systemd or another shell, so
the grep proved a FILE MENTIONS A NAME — never that the running daemon inherited it. The
Brain stayed quiet while ruflo recorded "opt-in required" every cycle.
A false ENABLED is strictly worse than a false disabled: nobody goes looking.
THREE STATES now, and only the process environment can prove the third:
off nothing set anywhere -> offer it
declared a settings file names it, but THIS -> fire, and say what is still missing
process did not inherit it
on this process HAS it truthy -> silent, genuinely opted in
THE GUIDANCE ALSO COULD NOT PRODUCE A WORKING SETUP. It named only the env var. A
downstream repo additionally requires a hash-pinned project-local anchor (ruflo #2840 /
PR #2848) or the runtime fails closed. "TURN IT ON" now names both steps, states BOTH data
gates (12 stored patterns to harvest, 20 held-out tasks to generate — the old copy quoted
only the first), and hands over the real check: `ruflo hooks intelligence --status`, read
the reason it prints.
Same shape as #136 and the lesson recorded from it: AUDITING THE WIRING IS NOT AUDITING THE
ANSWER. A mention is not a measurement.
ALSO — PR #137 (dependabot) was RED on windows-unit, and it was MY test, not the bump:
"exit 2 is the only code the host reads as a refusal: expected +0 to be 2". Every ADR-067
refusal policy is a .sh file; with no Git Bash on the runner resolveBash() returns nothing,
no policy contributes a verdict, and the gate correctly FAILS OPEN — the same behaviour
those four walls always had on a bashless host. The PRODUCT is right; the assertion demanded
a refusal that cannot occur there. The real-gate cases now require bash; the pure decide()
and policiesFor() cases carrying the precedence rule still run on every platform.
7/7 on the new suite, mutation-proved (restoring FLYWHEEL=on for the declared case fails the
TEETH case). 29/29 across the three affected suites.
Owner
|
@dependabot rebase |
dependabot
Bot
force-pushed
the
dependabot/npm_and_yarn/root-npm-ce52c755a9
branch
from
August 12, 2026 16:51
fc54b61 to
db05a22
Compare
dependabot
Bot
force-pushed
the
dependabot/npm_and_yarn/root-npm-ce52c755a9
branch
from
August 12, 2026 16:57
db05a22 to
c76a14a
Compare
Bumps the root-npm group with 3 updates in the / directory: @metaharness/flywheel, [@metaharness/router](https://github.com/ruvnet/metaharness) and [@metaharness/darwin](https://github.com/ruvnet/agent-harness-generator/tree/HEAD/packages/darwin-mode). Updates `@metaharness/flywheel` from 0.1.7 to 0.1.10 Updates `@metaharness/router` from 0.3.2 to 0.3.3 - [Release notes](https://github.com/ruvnet/metaharness/releases) - [Changelog](https://github.com/ruvnet/metaharness/blob/main/CHANGELOG.md) - [Commits](https://github.com/ruvnet/metaharness/commits) Updates `@metaharness/darwin` from 0.8.0 to 0.8.3 - [Release notes](https://github.com/ruvnet/agent-harness-generator/releases) - [Changelog](https://github.com/ruvnet/metaharness/blob/main/packages/darwin-mode/CHANGELOG.md) - [Commits](https://github.com/ruvnet/agent-harness-generator/commits/HEAD/packages/darwin-mode) --- updated-dependencies: - dependency-name: "@metaharness/darwin" dependency-version: 0.8.1 dependency-type: direct:development update-type: version-update:semver-patch dependency-group: root-npm - dependency-name: "@metaharness/flywheel" dependency-version: 0.1.8 dependency-type: direct:production update-type: version-update:semver-patch dependency-group: root-npm - dependency-name: "@metaharness/router" dependency-version: 0.3.3 dependency-type: direct:production update-type: version-update:semver-patch dependency-group: root-npm ... Signed-off-by: dependabot[bot] <support@github.com>
dependabot
Bot
force-pushed
the
dependabot/npm_and_yarn/root-npm-ce52c755a9
branch
from
August 12, 2026 17:04
c76a14a to
931dfe2
Compare
stuinfla
added a commit
that referenced
this pull request
Aug 13, 2026
…stead of reading
AgentDB had not durably persisted a write since 2026-08-10. `better_sqlite3.node` was
built for NODE_MODULE_VERSION 141 against a node needing 137, so the native bridge threw
ERR_DLOPEN_FAILED, ruflo fell back to sql.js, and three days of lessons, checkpoints and
session-end handoffs evaporated while the CLI printed `[OK] Data stored successfully`. The
warning printed on every one of those writes. I read it and continued — which is the actual
defect, and the reason none of this is fixed by printing it again.
degradation-watch.mjs A warning is text and text is skimmable, so this is a REFUSAL.
Registered in decision-gate (no new mechanism), it blocks
`git push` and `ruflo memory store` when the store cannot prove
it persists, and costs ~0ms on every other command. The prover
tests durability end-to-end — store, then confirm the row by SQL
— never the driver identity: grounding against ruflo's own source
showed sql.js is a legitimate path there, so a driver check would
cry wolf, and a channel that cries wolf gets skimmed.
brain-score.mjs Asked to grade the architecture I counted files, called it 52/100,
and reported it as quality while six graders sat unused. It had
measured CATALOGUE COVERAGE. `composite()` now REFUSES to average
across kinds, so that specific error cannot be made again, and
every dimension declares a freshness budget — `grounded 100/100`
was 34 days old and shipped on the README as "Current".
lesson-bridge.mjs Bridged rows were minted STATUS.RATIFIED because "the tag IS the
human act". But `--tags` runs in every session, from hooks and
from the model, and nightly-wrapper runs `--apply` unattended: any
process that could write a `lesson-*` row could inject standing
policy into every later session. I had myself hand-written a tagged
row that would have self-ratified. rUv's typed provenance (ADR-323)
already carried the answer eight lines up — `origin` honoured it and
`status` ignored it. Both now read the same fact. 20 of 43 drop to
candidate, each NAMED on stdout, because going quiet about 20
lessons no longer firing is the bigger failure.
resolveBash everywhere unprompted-runtime hardcoded `/bin/bash`, so on win32 spawnSync
errored, the fail-closed filter discarded every candidate, and the
ENTIRE unprompted plane — every lesson, every advocacy card — exited
0 with no diagnostic. The suite stayed green because its invariant
tests inject producers through a seam and never exercise the shipped
registry. Same literal in update-apply meant win32 SKIPPED the syntax
check on the hooks it was about to install. The resolver has existed
since #38 and sat one import away.
Swept as classes, not instances: a guard now fails on any unguarded `/bin/bash` spawn, and
the flywheel test that broke PR #137's windows-unit is fixed at its cause. Each guard is
proven by mutation — the detector fires on the exact line that shipped and clears the
correct forms — because "a test that cannot fail on broken code is not a test" is a lesson
this repo recorded and then violated again today.
Also lands kb/store-root.mjs: one resolver for the KB root, adopted by the retriever, with
a test that fails if a shipped path resolves from process.cwd() again.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Bumps the root-npm group with 3 updates in the / directory: @metaharness/flywheel, @metaharness/router and @metaharness/darwin.
Updates
@metaharness/flywheelfrom 0.1.7 to 0.1.10Updates
@metaharness/routerfrom 0.3.2 to 0.3.3Changelog
Sourced from @metaharness/router's changelog.
... (truncated)
Commits
Updates
@metaharness/darwinfrom 0.8.0 to 0.8.3Changelog
Sourced from @metaharness/darwin's changelog.
Commits