Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Catalog of all repo qv-* custom skills with one-line purpose and when-to-use. Use when the user asks what skills exist, which skill to use, how to invoke a skill, or invokes /qv-skill-list.
.claude/skills/tetherto-qv-skill-list/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 101% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 250% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 118% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 30% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 89% | 0% |
All custom repo skills live under .cursor/skills/ and use the qv- prefix. Invoke with /qv-<name> in chat.
Not listed here (by design):
setup — repo-wide bootstrap (/setup); copies shared config from packages/ocr-ggml/.agent/ into .claude/ and .cursor/.packages/ocr-ggml/.agent/skills/ — framework skills (orchestrate, release, ci-validate, commit-trace, review); installed by /setup, not qv-* prefixed.| Mode | Behavior | |------|----------| | Auto-invokable | No disable-model-invocation in frontmatter — the agent may load the skill when your request matches the description. | | Manual only | disable-model-invocation: true — invoke explicitly with /qv-<name> (or ask the agent to run that skill). |
When unsure which skill fits, scan the tables below or ask: "which qv skill should I use for …?"
| Skill | Purpose | Use when | |-------|---------|----------| | qv-pr-review | Deep PR review: gitflow, CI, format, code quality, security; posts a PENDING GitHub review with inline comments. | Reviewing any PR, given a PR URL, or follow-up from a pod status/my skill. Manual: /qv-pr-review | | qv-pr-test | Plan and run local PR validation; discovers packages/scripts, recommends test tier, analyzes results. | Testing a PR locally before merge. Manual: /qv-pr-test | | qv-pr-mine | Your open PRs across all pods, merge readiness, Slack ping messages for reviewers. | "What are my PRs?", merge readiness, who to ping. Manual: /qv-pr-mine |
| Skill | Purpose | Use when | |-------|---------|----------| | qv-devops-pr-create | PR titles and bodies for CI/CD, composite actions, automation, IaC. | Creating a DevOps PR. Manual: /qv-devops-pr-create | | qv-devops-pr-review | /qv-pr-review plus GitHub Actions security audit (pinning, permissions, OIDC, secrets). | Reviewing PRs that touch DevOps paths. Manual: /qv-devops-pr-review | | qv-devops-pr-status | Team DevOps PR dashboard: re-review, stale, needs-review, conflicts. | DevOps pod PR queue health. Manual: /qv-devops-pr-status | | qv-devops-why-my-pr-not | Diagnose missing CI checks or merge blockers (labels, CODEOWNERS, approvals). | "Why aren't checks running?" / "Why can't I merge?" Manual: /qv-devops-why-my-pr-not | | qv-devops-daily-update | Slack standup (Done / Planned / Blockers) from PRs, reviews, CI. | DevOps EOD or standup. Manual: /qv-devops-daily-update |
| Skill | Purpose | Use when | |-------|---------|----------| | qv-sdk-pr-create | PR descriptions for SDK pod packages (sdk, cli, rag, logging, error). | Creating an SDK pod PR; may chain to backmerge. Auto | | qv-sdk-pr-status | Team SDK pod PR dashboard. | SDK pod PR queue. Manual: /qv-sdk-pr-status | | qv-sdk-changelog | Changelogs for SDK pod packages (tag-based GitFlow). | Release prep, CHANGELOG_LLM.md. Auto | | qv-sdk-backmerge | Open backmerge PR (release version bump + changelog → main). | After SDK release PR; often chained from qv-sdk-pr-create. Auto | | qv-sdk-e2e-create | Scaffold e2e tests in packages/sdk/e2e for new/changed public APIs. | Adding consumer-facing SDK APIs. Auto | | qv-sdk-update-models | Regen SDK models.ts from live registry and open a [mod] PR; optional provider/python cascade. | After registry model updates; /qv-sdk-update-models. Manual | | qv-agent-stack-sync | Plan / draft-cascade / promote the SDK → CLI → AI SDK provider → OpenCode + OpenClaw release chain. | Agent-stack release prep; /qv-agent-stack-sync --plan. Auto |
| Skill | Purpose | Use when | |-------|---------|----------| | qv-addon-changelog | Changelog entries for a target add-on package. | Add-on release prep. Auto | | qv-addon-pr-create | PR descriptions for non-SDK inference addons, decoder, OCR. | "Prepare PR description" for an addon. Auto | | qv-addon-release-notes | Release notes for addon packages. | Addon release notes. Auto |
| Skill | Purpose | Use when | |-------|---------|----------| | qv-registry-autobase-patterns | Autobase, HyperDB, multi-writer patterns, replication debugging. | Registry server, Autobase/HyperDB schemas, Corestore issues. Auto |
| Skill | Purpose | Use when | |-------|---------|----------| | qv-dev-diary-setup | Opt-in local dev diary (~/.config/qvac-pr-skills/). | Enable diary, diary --init, configure capture. Manual: /qv-dev-diary-setup | | qv-dev-diary | Read, inspect, append diary entries. | View today's log, add notes after setup. Manual: /qv-dev-diary | | qv-daily-work-update | Personal EOD update from diary + GitHub + PR state (+ optional Asana). | Personal standup / "what did I do today". Manual: /qv-daily-work-update | | qv-asana-sync | Look up and update Asana tasks (status, comments, PR links). | Sync Asana task, move status, validate config. Manual: /qv-asana-sync |
| Skill | Purpose | Use when | |-------|---------|----------| | qv-holepunch-dev | Holepunch/P2P/Bare/Pear API discovery via docs.pears.com and gh. | Hypercore, Hyperswarm, Autobase, Bare, Pear development. Auto | | qv-notice-generate | NOTICE files and third-party attributions for monorepo packages. | License compliance, release NOTICE updates. Auto |
| Skill | Purpose | Use when | |-------|---------|----------| | qv-qip-triage | Conservatively checks whether a change clearly needs a QIP before deeper implementation. | Planning or implementing contract, cross-package, delivery, or principle-impacting work; PR review. Auto | | qv-qip-create | Drafts a Slack Canvas-ready QIP from a clear proposal or fuzzy idea. | Creating a QIP, shaping alternatives, consequences, and people to consult. Manual: /qv-qip-create | | qv-qip-review | Reviews a QIP for approval readiness, consequences, principle conflicts, and consultation coverage. | Reviewing a QIP draft before posting or approval. Auto |
Rule nudge: .cursor/rules/qip-triage.mdc
| You want to… | Skill | |--------------|-------| | Review any PR | qv-pr-review | | Review a DevOps/CI PR | qv-devops-pr-review | | Test a PR locally | qv-pr-test | | See your open PRs | qv-pr-mine | | SDK team PR board | qv-sdk-pr-status | | DevOps team PR board | qv-devops-pr-status | | Why CI/merge is blocked | qv-devops-why-my-pr-not | | Write SDK PR body | qv-sdk-pr-create | | Sync SDK models.ts from registry | qv-sdk-update-models | | Write addon PR body | qv-addon-pr-create | | SDK release changelog | qv-sdk-changelog | | Addon release changelog | qv-addon-changelog | | Backmerge after SDK release | qv-sdk-backmerge | | Sync CLI/provider/OpenCode/OpenClaw to SDK | qv-agent-stack-sync | | New SDK e2e tests | qv-sdk-e2e-create | | Registry Autobase/HyperDB help | qv-registry-autobase-patterns | | Holepunch stack help | qv-holepunch-dev | | Generate NOTICE files | qv-notice-generate | | Check if a QIP is needed | qv-qip-triage | | Draft a QIP | qv-qip-create | | Review a QIP | qv-qip-review | | List all skills | qv-skill-list (this file) |
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 13,997 | 6,801 | -51% | 1 | 1 | 0% | 2,343 | 3,902 | +67% | 0 | 0 | — |
case-02 | fail→pass | 16,283 | 26,576 | +63% | 1 | 1 | 0% | 2,746 | 5,518 | +101% | 0 | 0 | — |
case-03 | fail→pass | 5,555 | 4,278 | -23% | 1 | 1 | 0% | 967 | 3,386 | +250% | 0 | 0 | — |
case-04 | fail→pass | 9,073 | 3,462 | -62% | 1 | 1 | 0% | 1,499 | 3,261 | +118% | 0 | 0 | — |
case-05 | fail→pass | 14,980 | 3,131 | -79% | 1 | 1 | 0% | 2,482 | 3,239 | +30% | 0 | 0 | — |
case-06 | fail→fail | 23,915 | 3,998 | -83% | 1 | 1 | 0% | 1,980 | 3,373 | +70% | 0 | 0 | — |
case-07 | fail→pass | 12,150 | 3,172 | -74% | 1 | 1 | 0% | 1,716 | 3,240 | +89% | 0 | 0 | — |
case-08 | fail→pass | 9,900 | 2,719 | -73% | 1 | 1 | 0% | 1,547 | 3,130 | +102% | 0 | 0 | — |
case-09 | fail→pass | 11,043 | 2,277 | -79% | 1 | 1 | 0% | 1,757 | 3,044 | +73% | 0 | 0 | — |
case-10 | fail→pass | 10,901 | 3,163 | -71% | 1 | 1 | 0% | 1,629 | 3,201 | +97% | 0 | 0 | — |
case-11 | fail→pass | 12,821 | 3,385 | -74% | 1 | 1 | 0% | 2,109 | 3,157 | +50% | 0 | 0 | — |
case-12 | fail→pass | 7,584 | 3,234 | -57% | 1 | 1 | 0% | 1,174 | 3,239 | +176% | 0 | 0 | — |
case-13 | fail→pass | 8,709 | 3,784 | -57% | 1 | 1 | 0% | 1,114 | 3,263 | +193% | 0 | 0 | — |
case-14 | fail→pass | 10,250 | 4,538 | -56% | 1 | 1 | 0% | 1,312 | 3,296 | +151% | 0 | 0 | — |
case-15 | fail→pass | 10,612 | 2,695 | -75% | 1 | 1 | 0% | 1,385 | 3,057 | +121% | 0 | 0 | — |
case-16 | fail→pass | 12,412 | 2,497 | -80% | 1 | 1 | 0% | 1,711 | 3,073 | +80% | 0 | 0 | — |
case-17 | fail→pass | 8,588 | 2,924 | -66% | 1 | 1 | 0% | 1,447 | 3,021 | +109% | 0 | 0 | — |
case-18 | fail→pass | 12,974 | 3,420 | -74% | 1 | 1 | 0% | 1,738 | 3,085 | +78% | 0 | 0 | — |
case-19 | fail→pass | 15,309 | 4,984 | -67% | 1 | 1 | 0% | 2,245 | 3,338 | +49% | 0 | 0 | — |
case-20 | fail→pass | 18,365 | 3,270 | -82% | 1 | 1 | 0% | 2,565 | 3,246 | +27% | 0 | 0 | — |
case-21 | fail→pass | 13,051 | 3,518 | -73% | 1 | 1 | 0% | 1,799 | 3,220 | +79% | 0 | 0 | — |
case-22 | fail→pass | 11,891 | 2,712 | -77% | 1 | 1 | 0% | 1,555 | 3,090 | +99% | 0 | 0 | — |
case-23 | fail→pass | 11,889 | 2,895 | -76% | 1 | 1 | 0% | 1,743 | 3,015 | +73% | 0 | 0 | — |
case-24 | fail→pass | 10,055 | 3,735 | -63% | 1 | 1 | 0% | 1,300 | 3,175 | +144% | 0 | 0 | — |
case-25 | pass→pass | 19,980 | 16,060 | -20% | 1 | 1 | 0% | 1,650 | 3,818 | +131% | 0 | 0 | — |
case-26 | pass→pass | 15,891 | 12,265 | -23% | 1 | 1 | 0% | 2,299 | 4,333 | +88% | 0 | 0 | — |
case-27 | pass→pass | 10,405 | 9,382 | -10% | 1 | 1 | 0% | 1,567 | 4,288 | +174% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 27 cases were attempted. The headline lift of +81 percentage points is the difference between those two pass rates over the 27 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.