OmniRoute CLI Evals
diegosouzapw/OmniRoute
Creates and runs LLM evaluation suites from the omniroute CLI, follows live runs, shows scorecards, compares models and ties eval runs into CI.
Release the qvac-fabric consumers to npm after the bundled version-bump PR merges — one release branch + on-merge dispatch per consumer, stopping at the npm approval gate.
The automated check flagged lines worth reading first. See the safety section below.
$ npx skills add tetherto/qvac --skill release-fabric-consumers -a claude-codeProject install by default; add -g for ~/.claude/skills/.
$ gh skill install tetherto/qvac release-fabric-consumers --agent claude-codeProject scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).
$ git clone --depth 1 https://github.com/tetherto/qvac.git skills-src && mkdir -p .claude/skills && cp -r skills-src/packages/ocr-ggml/.agent/skills/release-fabric-consumers .claude/skills/release-fabric-consumers && rm -rf skills-srcUse ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.
Claude Code skills documentation · loads skills from .claude/skills/
Install the "release-fabric-consumers" agent skill from https://github.com/tetherto/qvac/tree/main/packages/ocr-ggml/.agent/skills/release-fabric-consumers into .claude/skills/release-fabric-consumers/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "release-fabric-consumers", then confirm the skill loads.Claude Code copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$skill-installer install https://github.com/tetherto/qvac/tree/main/packages/ocr-ggml/.agent/skills/release-fabric-consumersType this inside Codex. $skill-installer <name> installs a curated skill from openai/skills. The installer writes to $CODEX_HOME/skills (default ~/.codex/skills). Restart Codex if the skill does not show up.
$ npx skills add tetherto/qvac --skill release-fabric-consumers -a codexProject install goes to .agents/skills/; add -g for ~/.codex/skills/.
$ gh skill install tetherto/qvac release-fabric-consumers --agent codexProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/tetherto/qvac.git skills-src && mkdir -p .agents/skills && cp -r skills-src/packages/ocr-ggml/.agent/skills/release-fabric-consumers .agents/skills/release-fabric-consumers && rm -rf skills-srcUse ~/.agents/skills/ instead of .agents/skills for a personal install.
Codex skills documentation · loads skills from .agents/skills/
Install the "release-fabric-consumers" agent skill from https://github.com/tetherto/qvac/tree/main/packages/ocr-ggml/.agent/skills/release-fabric-consumers into .agents/skills/release-fabric-consumers/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "release-fabric-consumers", then confirm the skill loads.Codex copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ npx skills add tetherto/qvac --skill release-fabric-consumers -a cursorProject install goes to .agents/skills/; add -g for ~/.cursor/skills/.
$ gh skill install tetherto/qvac release-fabric-consumers --agent cursorProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/tetherto/qvac.git skills-src && mkdir -p .cursor/skills && cp -r skills-src/packages/ocr-ggml/.agent/skills/release-fabric-consumers .cursor/skills/release-fabric-consumers && rm -rf skills-srcUse ~/.cursor/skills/ instead of .cursor/skills for a personal install.
Cursor skills documentation · loads skills from .cursor/skills/, .agents/skills/, .claude/skills/, .codex/skills/
Install the "release-fabric-consumers" agent skill from https://github.com/tetherto/qvac/tree/main/packages/ocr-ggml/.agent/skills/release-fabric-consumers into .cursor/skills/release-fabric-consumers/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "release-fabric-consumers", then confirm the skill loads.Cursor copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ gemini skills install https://github.com/tetherto/qvac.git --path packages/ocr-ggml/.agent/skills/release-fabric-consumers--scope user (default) or --scope workspace; --path is the subfolder of the repo that holds the skill; --consent skips the security confirmation prompt.
$ npx skills add tetherto/qvac --skill release-fabric-consumers -a gemini-cliProject install goes to .agents/skills/; add -g for ~/.gemini/skills/.
$ gh skill install tetherto/qvac release-fabric-consumers --agent gemini-cliProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/tetherto/qvac.git skills-src && mkdir -p .gemini/skills && cp -r skills-src/packages/ocr-ggml/.agent/skills/release-fabric-consumers .gemini/skills/release-fabric-consumers && rm -rf skills-srcUse ~/.gemini/skills/ instead of .gemini/skills for a personal install, then run /skills reload.
Gemini CLI skills documentation · loads skills from .gemini/skills/, .agents/skills/
Install the "release-fabric-consumers" agent skill from https://github.com/tetherto/qvac/tree/main/packages/ocr-ggml/.agent/skills/release-fabric-consumers into .gemini/skills/release-fabric-consumers/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "release-fabric-consumers", then confirm the skill loads.Gemini CLI copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ gh skill install tetherto/qvac release-fabric-consumersInstalls for Copilot at project scope by default; add --scope user for a personal install. Preview a skill first with gh skill preview. Needs GitHub CLI 2.90.0 or later (public preview).
$ npx skills add tetherto/qvac --skill release-fabric-consumers -a github-copilotProject install goes to .agents/skills/; add -g for ~/.copilot/skills/.
$ git clone --depth 1 https://github.com/tetherto/qvac.git skills-src && mkdir -p .github/skills && cp -r skills-src/packages/ocr-ggml/.agent/skills/release-fabric-consumers .github/skills/release-fabric-consumers && rm -rf skills-srcUse ~/.copilot/skills/ instead of .github/skills for a personal install. Commit .github/skills so cloud agent and code review can use it.
GitHub Copilot skills documentation · loads skills from .github/skills/, .claude/skills/, .agents/skills/
Install the "release-fabric-consumers" agent skill from https://github.com/tetherto/qvac/tree/main/packages/ocr-ggml/.agent/skills/release-fabric-consumers into .github/skills/release-fabric-consumers/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "release-fabric-consumers", then confirm the skill loads.GitHub Copilot copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ npx skills add tetherto/qvac --skill release-fabric-consumers -a opencodeOpenCode documents no install command of its own. Project install goes to .agents/skills/; add -g for ~/.config/opencode/skills/.
$ gh skill install tetherto/qvac release-fabric-consumers --agent opencodeProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/tetherto/qvac.git skills-src && mkdir -p .opencode/skills && cp -r skills-src/packages/ocr-ggml/.agent/skills/release-fabric-consumers .opencode/skills/release-fabric-consumers && rm -rf skills-srcUse ~/.config/opencode/skills/ instead of .opencode/skills for a personal install.
OpenCode skills documentation · loads skills from .opencode/skills/, .claude/skills/, .agents/skills/
Install the "release-fabric-consumers" agent skill from https://github.com/tetherto/qvac/tree/main/packages/ocr-ggml/.agent/skills/release-fabric-consumers into .opencode/skills/release-fabric-consumers/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "release-fabric-consumers", then confirm the skill loads.OpenCode copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
release-fabric-consumersRelease the qvac-fabric consumers to npm after the bundled version-bump PR merges — one release branch + on-merge dispatch per consumer, stopping at the npm approval gate.
Release Fabric Consumers is an agent skill from tetherto/qvac. Release the qvac-fabric consumers to npm after the bundled version-bump PR merges — one release branch + on-merge dispatch per consumer, stopping at the npm approval gate. Supports --exclude to hold packages back. Phase C after rollout-phase-b.
Its SKILL.md is about 4.3k tokens, which your agent loads only when the skill is triggered. It is a single SKILL.md file with no bundled scripts.
It sits in AI & LLM Engineering. It works with npm. The repository describes itself as: Open-source local AI SDK - run AI on-device with no cloud, no API keys. Supports GGUF, RAG, image, music, and video generation, speech-to-text, P2P inference, and more… The licence is Apache-2.0.
6 steps, taken from the step headings in SKILL.md.
Read from SKILL.md and the folder at commit f874b3a. It shows what the files ask for, not the result of running them.
Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.
From allowed-tools in the SKILL.md frontmatter.
Shell commands in SKILL.md call:
gitghnpmbashFrom the folder's file list and the shell code blocks in SKILL.md.
No URLs in SKILL.md. Its commands use git, gh and npm, which can reach the network depending on how they are called.
From URLs in SKILL.md, links to its own repository left out.
Names these keys or tokens, usually read from environment variables:
NPM_TOKENFrom names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.
Release Fabric Consumers loads about 4.3k tokens when it runs. Until then it costs about 67 tokens; SKILL.md has 2,083 words of instructions outside code blocks.
Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.
The automated check found patterns that need a careful read before installing.
bare registry `.npmrc` and an **isolated `HOME`** (so the user's token is never read),Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.
The full file from tetherto/qvac at commit f874b3a, republished under its Apache-2.0 licence (© tetherto). 2,083 words, ~4,299 tokens.
.claude/skills/release-fabric-consumers/SKILL.md (or your agent's skills folder).Publish the qvac-fabric consumers to npm in one coordinated pass. This is the
final phase of a fabric rollout — the step after rollout-phase-b's bundled
version-bump PR (vcpkg version>= + package.json + CHANGELOG.md for the full roster)
has merged to main. It applies the single-package release workflow to every consumer
in the release set at once.
It reuses the same mechanics as /release (cut a release-* branch → dispatch
on-merge-nx.yml with the package as an input → the on-merge workflow builds prebuilds,
publishes to npm with the latest tag, and creates a git tag), but for the whole release
set, and it is aware of the per-consumer quirks and the mandatory human approval gate.
The release set is the full consumer roster (see the table below) minus anything named
in --exclude. Every step below operates on the release set, not the roster.
/release-fabric-consumers [base-branch] [--exclude <pkg>[,<pkg>...]]
/release-fabric-consumers
/release-fabric-consumers main --exclude model-fit
/release-fabric-consumers main --exclude model-fit,vla-ggml[base-branch] — optional; the branch the version bumps landed on. Defaults to main.--exclude <pkg>[,<pkg>...] — optional; comma-separated packages/ directory names
(the first column of the consumer table) to hold back from this pass. Everything else in
the roster is released.Validate --exclude before touching anything. A value that does not match a packages/
dir name in the table is a hard error, not a silent no-op — a typo like
--exclude modelfit would otherwise quietly release the very package the user meant to hold
back. Report the unmatched name and stop before creating any release branch. Excluding
every consumer is likewise an error, not an empty success.
<base-branch> — either rollout-phase-b's own bundled PR
(e.g. qvac#3334) or, in --on-top-of-pr mode, the feature PR the bumps rode in on (e.g.
qvac#3725). The target versions live on origin/<base-branch>.origin/<base-branch>'s package.json version is
higher than the current npm latest (i.e. npm is one bump behind). If not, the version
bump PR hasn't merged yet — stop.CHANGELOG.md has a ## [<version>] - <date> section (bracketed
heading — the release-merge-guard requires this exact format).These are checked only for the release set — a package held back by --exclude is never inspected,
which is the point of excluding it.
The "npm is one bump behind" check can pass for the wrong reason. A package whose npm latest
reads 0.0.0 with no release tag has never actually been published — 0.0.0 is a name placeholder,
so any real version compares as "higher" and the precondition passes while the package is in a
first-publish state this skill otherwise assumes away.
npm view @qvac/<pkg> dist-tags.latest
git -C <repo> ls-remote --tags origin "<tag-prefix>-v*"Both signals together (0.0.0 and no tag) mean a first release. 0.0.0 here is a sentinel,
not a version that drifts — it is npm's placeholder for a name reserved but never published, so it
stays literal. Which package is in that state does drift, so run the two commands above across the
roster rather than trusting a name written here; model-fit was the instance when this was written
and has since published. When a package is in that state, its first release needs two things settled
before dispatch:
## [<version>] - <date> entry — a CHANGELOG.md holding only an ## [Unreleased]
block is what the release extractor will not accept as a version section;origin/main publishes as-is, or is superseded by
the rollout bump. Do not bump past an unpublished version without asking — that silently burns a
version number.Until both are settled, hold it back with --exclude <pkg> rather than dispatching it and
dealing with a failed release-merge-guard. That is the supported way to release the rest of the
roster on schedule while a first-publish package waits.
This table is the full roster. The release set is this list minus --exclude.
Package dir (packages/) | npm name | on-merge workflow | git-tag created |
|---|---|---|---|
embed-llamacpp | @qvac/embed-llamacpp | on-merge-nx.yml | llamacpp-embed-v<ver> |
fabric | @qvac/fabric | on-merge-nx.yml | fabric-v<ver> |
llm-llamacpp | @qvac/llm-llamacpp | on-merge-nx.yml | llamacpp-llm-v<ver> |
model-fit | @qvac/model-fit | on-merge-nx.yml | model-fit-v<ver> |
ocr-ggml | @qvac/ocr-ggml | on-merge-nx.yml | ocr-ggml-v<ver> |
translation-nmtcpp | @qvac/translation-nmtcpp | on-merge-nx.yml | v<ver> (bare) |
vla-ggml | @qvac/vla-ggml | on-merge-nx.yml | vla-v<ver> |
The git-tag column is not uniform. Use the package's entry when checking a release.
classification-ggml is not in this list. It dropped the qvac-fabric vcpkg
dependency and now consumes the published npm package @qvac/fabric, so it is not part
of a fabric rollout — see Release fabric first below for the one follow-up it needs.
git/gh operations in the code working repo (e.g. ~/repo/qvac) — NOT a
stale/read-only worktree. Verify the repo is on the right remote (tetherto/qvac).origin/<base-branch>, never on local HEAD (local
main is often behind).origin/<base-branch>. Create it without a checkout, then assert an empty diff before
pushing. Never edit anything under .github/workflows/**. If the diff is non-empty,
STOP and report — do not push.Do the whole release set in parallel unless told otherwise (version-only bumps were already build-validated on the fabric sync PR; a canary-first order is a fine alternative).
fabric first@qvac/fabric is the shared runtime addon the others build against, and it is also a
caret dependency of @qvac/classification-ggml. Publish it ahead of the rest so
downstream installs resolve against the new build rather than the previous one.
Then check the caret. Read both values rather than assuming either — they move every release, so any version written here would be wrong by the time you read it:
git -C <repo> grep -h "@qvac/fabric" origin/main -- packages/classification-ggml/package.json
git -C <repo> grep -h '"version"' origin/main -- packages/fabric/package.jsonOn a 0.x version a caret locks the minor: ^0.<m>.<p> absorbs later 0.<m>.x
automatically but will not cross to 0.<m+1>.0. So if this release moves
@qvac/fabric's minor past the one classification-ggml's caret pins,
classification-ggml needs a manual dependency bump — track it as a follow-up after the
release, not as part of the rollout PR. This has happened before and is expected to
recur; packages/classification-ggml/CHANGELOG.md records each crossing.
If fabric is in --exclude, this ordering step does not apply to that pass and the
classification-ggml caret follow-up does not arise — release the rest of the set in
parallel and say in the report that the ordering constraint was moot.
git -C <repo> fetch origin <base-branch>Resolve the release set first (roster minus --exclude, with --exclude validated per Usage),
then for each consumer in the release set read the version off origin/<base-branch>
(git -C <repo> show origin/<base-branch>:packages/<pkg>/package.json) and confirm
npm view @qvac/<pkg> version is lower. Also confirm no target release-* branch already
exists on the remote (git -C <repo> ls-remote --heads origin "release-<pkg>-<ver>").
State the resolved release set and the excluded packages before proceeding, so the user can catch a wrong set before any branch is pushed.
git -C <repo> branch release-<pkg>-<ver> origin/<base-branch> # no checkout
git -C <repo> diff --stat origin/<base-branch>..release-<pkg>-<ver> # MUST be empty
git -C <repo> push origin release-<pkg>-<ver> # non-force, non-maingh workflow run on-merge-nx.yml --repo tetherto/qvac \
--ref release-<pkg>-<ver> -f package=<pkg>package is a required choice input and takes the package directory name, the same
string the branch carries.
Pushing a fresh release-* branch usually does not auto-trigger (path filter sees no
new commits), so the explicit dispatch is the trigger — but check
gh run list --branch release-<pkg>-<ver> and, if a push-triggered run already exists, do
NOT double-dispatch.
Poll until each run pauses at publish-npm. Read the run URL straight from gh rather than
hand-assembling it from a run id:
gh run list --branch release-<pkg>-<ver> --repo tetherto/qvac --limit 1 --json databaseId,url,status,conclusion
gh api repos/tetherto/qvac/actions/runs/<run-id>/pending_deploymentsJob order per run: label-gate (auto) → publish-logic + release-merge-guard (env
release, branch-policy only) → build (prebuilds, ~9 platforms) → publish-npm (env
npm, required reviewers) ← PAUSES HERE → create-tag.
When a run reaches the npm pending-deployment gate, STOP and hand off to the user.
Required output format. Emit the CI run links as one line per addon, in the order of the consumer table, and nothing else on those lines:
<addon>: <CI run link><addon> is the packages/ directory name (fabric, embed-llamacpp, vla-ggml,
translation-nmtcpp, …) — not the npm name, not the workflow name.<CI run link> is the run's url from the gh run list --json call above, i.e.
https://github.com/tetherto/qvac/actions/runs/<run-id>.--exclude get no line.— awaiting gate, — build running, — failed) so the list is never
partial and a missing package is always a real problem rather than a timing artifact.Worked example (--exclude model-fit):
fabric: https://github.com/tetherto/qvac/actions/runs/31390432306
embed-llamacpp: https://github.com/tetherto/qvac/actions/runs/31390431726
llm-llamacpp: https://github.com/tetherto/qvac/actions/runs/31390432200
ocr-ggml: https://github.com/tetherto/qvac/actions/runs/31390430929
translation-nmtcpp: https://github.com/tetherto/qvac/actions/runs/31390430912
vla-ggml: https://github.com/tetherto/qvac/actions/runs/31390432221 — build runningAfter the list, note the gate detail once — gated job publish-npm, environment npm, and which
runs are already pending approval. That is context; the list is the hand-off artifact.
Never approve the deployment yourself (the automation token has can_approve=false anyway;
only a qvac-internal-release member can).
Once publish-npm + create-tag complete for a consumer:
npm view @qvac/<pkg> dist-tags.latest # == <ver>
git -C <repo> ls-remote --tags origin "<git-tag>" # tag exists (see per-package table)Public-access check (token-less): install each @qvac/<pkg>@<ver> into a clean dir with a
bare registry .npmrc and an isolated HOME (so the user's token is never read),
--ignore-scripts --no-package-lock; confirm the installed package.json version matches.
Put this loop in a script file and run bash <file> (see Bash discipline).
Consolidated table: package, npm latest, git tag, public-install ✓, CI run link. Release
branches are never merged back to main; offer to delete the leftover local branches.
State explicitly which packages were excluded and why (e.g. model-fit — first publish, awaiting dated CHANGELOG entry and version decision). A pass that silently covers six of seven is
indistinguishable later from one that dropped a package by accident.
on-merge-nx.yml, selected by the package input.
Both the input and the release branch use the package directory name, including
release-vla-ggml-<ver>.detect parses release-<pkg>-<x.y.z>,
requires packages/<pkg>/project.json to exist, and refuses to publish unless the run
selected exactly that package. A misnamed branch fails the run rather than publishing
the wrong thing.<pkg>-v<ver>. Use the table:
llamacpp-embed-v<ver>, fabric-v<ver>, llamacpp-llm-v<ver>, model-fit-v<ver>,
ocr-ggml-v<ver>, bare v<ver> for translation-nmtcpp, and vla-v<ver>.
The tag comes from create-release-tag.yml, which builds <repo_name>-v<published_version>
from repoName in each package's project.json (targets.on-merge.options.ci) — read
that rather than guessing from the directory name if a new consumer appears.label-gate/Authorise job treats
push and workflow_dispatch as trusted events (authorised=true) — so a
release-* branch dispatch needs no labels at all. PR events are the contrasting
case, and they do not want verified either: they need the granular stage labels
(prebuilds plus the four run-*), which are the only names ci-router reads.
verified selects nothing anywhere — the label gate was retired repo-wide.publish-npm runs in the npm
environment, protected by required_reviewers (team qvac-internal-release). Every run
pauses there. Detect via .../pending_deployments. Never approve — hand off to the
user. (Earlier jobs use env release, which has only a branch_policy allowing
release-*, so they never gate.)gh release view <tag> returns "release not found" — that is EXPECTED, not a
failure. (The create-github-release.yml workflow is unrelated/defunct on another
branch.) Do not treat a missing GitHub Release as an error.main; local copies remain after the run —
offer to clean them up. If a target release-* branch already exists on the remote, ask
the user before reusing or aborting.CLAUDE.md). In direct commands: no heredocs, no $()
command substitution, no &&/||/; chaining, no pipes or redirects. Put polling
loops / the public-install check in a script file and run bash <file>. Use git -C <path> and gh --repo <repo> instead of cd.release-merge-guard fails → version not bumped or CHANGELOG.md missing the
## [<version>] - <date> section. Fix on main (new PR) and re-dispatch.gh run view <run-id> --repo tetherto/qvac --log-failed.NPM_TOKEN
is invalid; check npm view @qvac/<pkg> versions.qvac#3334 (qvac-fabric 9840.0.0) released all 6 consumers of that time via this exact
flow. Runs: classification-ggml 29736362703, embed-llamacpp 29736364449,
llm-llamacpp 29736366470, ocr-ggml 29736373159, translation-nmtcpp 29736374927,
vla-ggml 29736376662. All paused at publish-npm; after a qvac-internal-release member
approved, all 6 published to npm latest and created their (differently-named) git tags.
This is a historical record, not the current roster — classification-ggml was still a
fabric consumer then, and fabric / model-fit had not been added. The flow it demonstrates
is unchanged; use the table above for who to release.
© tetherto, Apache-2.0. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file
Just SKILL.md in packages/ocr-ggml/.agent/skills/release-fabric-consumers of tetherto/qvac.
Open the folder on GitHubat commit f874b3a
Release Fabric Consumers next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.
| Skill | Stars | Used in | Tokens | Auto-check | Licence | Repo updated |
|---|---|---|---|---|---|---|
| Release Fabric Consumers this skilltetherto/qvac | 683 | — | ~4.3k | Automated safety check: Warn | Apache-2.0 | |
| OmniRoute CLI Evalsdiegosouzapw/OmniRoute | 74k | — | ~1.3k | Automated safety check: Pass | MIT | |
| Code Change Verificationopenai/openai-guardrails-js | 105 | — | ~619 | Automated safety check: Pass | MIT | |
| Openclaw SetupFerroxLabs/wayland | 608 | — | ~1.9k | Automated safety check: Pass | AGPL-3.0 | |
| Skill HubCoWork-OS/CoWork-OS | 474 | — | ~586 | Automated safety check: Pass | MIT | |
| Compromise NLP for JavaScriptspencermountain/compromise | 12k | — | ~1.8k | Automated safety check: Pass | MIT |
diegosouzapw/OmniRoute
Creates and runs LLM evaluation suites from the omniroute CLI, follows live runs, shows scorecards, compares models and ties eval runs into CI.
openai/openai-guardrails-js
Select and run Guardrails verification for code, dependency, test, tooling, documentation, or repository-workflow changes.
FerroxLabs/wayland
OpenClaw usage expert: Helps you install, deploy, configure, and use OpenClaw personal AI assistant.
CoWork-OS/CoWork-OS
Use the skill-hub CLI to search, install, update, and publish agent skills from skill-hub.com.
spencermountain/compromise
Guide for writing correct code with the compromise rule-based NLP library: tagging, match syntax, in-place transforms and common tasks like tense changes and redaction.
andrewyng/context-hub
Guides building Tavily integrations for web search, URL extraction, site crawling and AI-assisted research in Python or JavaScript agent and RAG projects.
tetherto/qvac
Creates a Solutions page in the QVAC documentation website from a real use case, generalizing the case into reusable guidance and registering the page in the site navigation.
tetherto/qvac
Updates the docs website after a change to the SDK or CLI. An agent skill from tetherto/qvac.
tetherto/qvac
Plan and prepare the QVAC agent-stack release cascade across @qvac/inference, @qvac/sdk, @qvac/cli, @qvac/ai-sdk-provider, @qvac/opencode-plugin, and @qvac/openclaw-plugin.
tetherto/qvac
Run the deterministic code-quality audit, turn related findings into contextual remediation groups, prepare approval-gated Asana proposals, reconcile recurring runs, or configure twice-monthly…
tetherto/qvac
Review C++ changes for string parameter and call-site efficiency conventions (std::stringview, std::string&&, const std::string&, const char, and TransparentStringMap lookup).
tetherto/qvac
Generate changelog entries for a target add-on package. An agent skill from tetherto/qvac.
Works with
Categories
Release the qvac-fabric consumers to npm after the bundled version-bump PR merges — one release branch + on-merge dispatch per consumer, stopping at the npm approval gate. Release Fabric Consumers is an agent skill from tetherto/qvac. Release the qvac-fabric consumers to npm after the bundled version-bump PR merges — one release branch + on-merge dispatch per consumer, stopping at the npm approval gate.
Release Fabric Consumers fits situations like: AI & LLM Engineering work in your project.
Run `npx skills add tetherto/qvac --skill release-fabric-consumers -a claude-code`. Or copy the skill folder (packages/ocr-ggml/.agent/skills/release-fabric-consumers in tetherto/qvac) into .claude/skills/release-fabric-consumers in your project. Claude Code loads it when a task matches its description.
Run `npx skills add tetherto/qvac --skill release-fabric-consumers -a codex`. Or copy the skill folder (packages/ocr-ggml/.agent/skills/release-fabric-consumers in tetherto/qvac) into .agents/skills/release-fabric-consumers in your project. Codex loads it when a task matches its description.
Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add tetherto/qvac --skill release-fabric-consumers -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/release-fabric-consumers, .gemini/skills/release-fabric-consumers, .github/skills/release-fabric-consumers and .opencode/skills/release-fabric-consumers in your project.
Going by SKILL.md and its folder, Release Fabric Consumers needs the command-line tools its instructions call (git, gh, npm and bash) and credentials named NPM_TOKEN.
SKILL.md contains no URLs. Its commands use git, gh and npm, which can reach the network depending on how they are called. This is read from the text; nothing was executed.
Our automated static check of SKILL.md flagged 1 warning(s): mentions a credentials file (ssh keys, cloud or package-manager tokens). Read the flagged lines before installing; the check is not a guarantee either way.
Release Fabric Consumers is published under the Apache-2.0 licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.
About 4.3k tokens (SKILL.md is roughly 17k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full.
Skills that share tags, products or a category with Release Fabric Consumers: OmniRoute CLI Evals (diegosouzapw/OmniRoute, 74k stars), Code Change Verification (openai/openai-guardrails-js, 105 stars), Openclaw Setup (FerroxLabs/wayland, 608 stars) and Skill Hub (CoWork-OS/CoWork-OS, 474 stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.
tetherto (a GitHub organization) maintains it in tetherto/qvac, which has 683 GitHub stars. The repository holds 50 skills in this directory. The repository was last updated on October 9, 2026.
Source: tetherto/qvac on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.