Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add YuDefine/nuxt-supabase-starter --skill dep-upgradegit clone --depth 1 https://github.com/YuDefine/nuxt-supabase-starterWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/yudefine/nuxt-supabase-starter/dep-upgrade)<a href="https://agentmods.dev/skills/yudefine/nuxt-supabase-starter/dep-upgrade"><img src="https://agentmods.dev/badge/skills/yudefine/nuxt-supabase-starter/dep-upgrade/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/yudefine/nuxt-supabase-starter/dep-upgrade"><img src="https://agentmods.dev/badge/skills/yudefine/nuxt-supabase-starter/dep-upgrade.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00030 | $0.04758 |
| Opus 5 | $0.00015 | $0.02379 |
| Sonnet 5 | $0.00006 | $0.00952 |
| Haiku 4.5 | $0.00003 | $0.00476 |
Grade C, and why
dep-upgrade scanned grade C with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Hidden instructionshighPrompt injection
Directives inside HTML comments, invisible characters or bidirectional overrides are read by the model and not by the person reviewing the file.
``` PHASE_RESULT: SUCCESS COMMIT: <sha> FILES_CHANGED: package.json, <lockfile> VERIFICATION: <依驗證步驟回報> ``` 失敗： ``` PHASE_RESULT: FAILURE STAGE_FAILED: <install | typecheck | build | test> ERROR_TAIL: <≤ 30 行 error How it starts
The opening of the file, as written. The whole thing — 334 lines — stays where its author put it; the contents beside it link to each section on GitHub.
dep-upgrade — 統一 dep 升級入口
三種升級需求一個 skill 涵蓋：
- Outdated batch：一個 consumer 累積很久沒升、
pnpm outdated一坨要批升 - Fleet sweep：看到 upstream release、想跨 consumer 跟上同一個版本
- Skills：
npx skills add裝進來的第三方 skill 落後上游，或上游新增了我們還沒裝的 skill
skill 開頭依輸入分流，不要記三個 skill 名。
Step 0 — Mode dispatcher（最先讀）
依輸入跟 cwd 分流：
| 觸發 | cwd 預期 | Mode | 跳到 |
|---|---|---|---|
純命令 /dep-upgrade（無參數） |
consumer root（有 package.json + lockfile 或 .github/workflows/） |
Outdated batch | § Outdated · Step O.1 |
/dep-upgrade actions |
consumer root（有 .github/workflows/） |
Outdated batch（only actions） | § Outdated · Step O.1（skip O.1-npm） |
給 GitHub release URL（https://github.com/.../releases/tag/v<ver>） |
clade home | Fleet | § Fleet · Step F.1 |
給 <pkg>@<ver> / <pkg> v<ver> / <pkg> <ver> |
clade home | Fleet | § Fleet · Step F.1 |
| 給純 pkg name（「升 @nuxt/ui」、「無腦升 X」） | clade home | Fleet + Discovery | § Fleet · Step F.1 |
| 給 package manager / runtime 本身（「升 pnpm」、「fleet packageManager 統一」、「Node 版本統一」） | clade home | Fleet · Toolchain | § Fleet · Step F.1，再讀 fleet-mode.md § Toolchain sweep 分支 |
/dep-upgrade skills，或提到 skill 上游 / 落後 / 新增（「supabase skill 上游更新了我們有跟嗎」「掃一下 skill 有沒有落後」） |
clade home | Skills | § Skills · Step S.1 |
| 無參數但 cwd = clade home | — | STOP + 問意圖 | 見下方 § Disambiguation |
Disambiguation（cwd 與 input 不對）
- 在 consumer 給 pkg name → 問「你想 (A) 只對這個 consumer 升 (走 Outdated batch 但鎖單套件)、還是 (B) 跨 registry 全命中 consumer sweep (要
cd ~/offline/clade再跑)？」 - 在 clade 無參數 → 問「你想 (A) sweep 哪個 pkg？(B) 進某個 consumer 跑
pnpm outdated？還是 (C) 掃第三方 skill 的上游落後（Skills mode）？」 - 在 worktree（cwd 含
-wt/） + 無參數 → 視為已在 Outdated batch 中段，跳 Step O.0、直接續跑 Step O.2
MUST 等 user 拍板，NEVER 主線自選 mode。
共用基礎（兩個 mode 都用到）
| 基礎 | 出處 |
|---|---|
| Worktree gate | [[worktree-default]] §1，wt-helper 開 / merge-back。機械檢查點見下方 § Worktree gate（fail-closed） |
| Pi 派工模板 | vendor/snippets/pi-upgrade-prompts/{first-pass,research}.md（authoring source），SKILL.md § Pi prompt templates 是 plugin cache 副本 |
| Pi watch protocol | [[agent-routing.pi-watch-protocol]] |
| Selective stage on main | 一律 git add package.json <lockfile> + 額外指定檔，NEVER git add -A |
| Commit msg（commitlint-aware） | subagent / pi 端讀 consumer commitlint config 後生 compliant msg |
What ships with it
3 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- yesterday Changed b4faa5b20fca
- 4d ago Changed · +4 lines 16297d502de1
- 8d ago First seen · 330 lines · 30 tokens per session scan C 162bb1829017
dep-upgrade is a skill published in the GitHub repository YuDefine/nuxt-supabase-starter (45 stars, last pushed today), licensed MIT. It adds 30 tokens to every session and 4,758 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it C with 1 finding (hidden instructions). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-03.
Other skills, from other repositories
gh-issue
Size-audit, write, and split BanyanDB issues that somebody else or an automated TDD workflow can implement. Use whenever the user asks to file or revise an issue, decide whether an issue is too large, make an issue TDD-ready, turn a design into tickets, or split an umbrella into executable leaves. Do not draft or file…
nw-finalize
Archives a completed feature to docs/evolution/, migrates lasting artifacts to permanent directories, and cleans up the temporary workspace. Use after all implementation steps pass and mutation testing completes.
linear-maintenance
Bi-weekly housekeeping pass over the evlog team's Linear backlog — labels, projects, priorities, stale triage, work that is done but still open, and anything misfiled next to its siblings. Applies the mechanically verifiable fixes and surfaces the rest as decisions. Load this when the linear-maintenance schedule…
nw-new
Guided wizard to start a new feature. Asks what you want to build, recommends the right starting wave, and launches it.
nw-deliver-orchestration
DELIVER wave orchestration workflow -- 9 phases from baseline to finalization. Load when user invokes deliver command. Covers state tracking, smart skip logic, retry, resume, and quality gate enforcement.
prd-v07-test-planning
Define test cases BEFORE implementation, ensuring every API, business rule, and user journey has verifiable acceptance criteria during PRD v0.7 Build Execution. Triggers on requests to define tests, plan test coverage, create test cases, or when user asks "define tests", "test planning", "what to test?", "test cases"…