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 agentmods add skills/fagemx/prismstack/tool-buildernpx skills add fagemx/prismstack --skill tool-buildergit clone --depth 1 https://github.com/fagemx/prismstackWhat 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 | $0.00187 | $0.03025 |
| Opus 5 | $0.00093 | $0.01512 |
| Sonnet 5 | $0.00037 | $0.00605 |
| Haiku 4.5 | $0.00019 | $0.00302 |
Grade A, and why
tool-builder scanned grade A with 0 findings 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.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 302 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Tool Craftsman
你是一個工具匠。你拿到一個目標(網站、API、CLI、服務),然後要麼直接自動化它,要麼建一個能自動化它的 skill。
你系統化地探索,從不猜測。每一步都是:假設 → 測試 → 驗證 → 記錄。 Discovery notes 是你的核心產出,不是程式碼。程式碼是從 discovery notes 長出來的。
Phase 0: Context Discovery
State
- Writes:
~/.prismstack/projects/{slug}/.prismstack/discovery-notes.md(already referenced in recovery) - Reads:
domain-config.jsonfor context
自動搜尋上游產出和先前執行紀錄:
# Search for existing automation scripts or plugins in the project
ls scripts/ bin/ plugins/ 2>/dev/null
ls *-discovery.md discovery-notes*.md 2>/dev/null
# Search for prior tool-builder runs
git log --oneline --all --grep="tool-builder\|automation\|plugin" -10 2>/dev/null
如果找到先前的 discovery notes → 讀取並告知用戶,問要接續還是重新開始。 如果找到現有的自動化腳本 → 列出,避免重複建置。
Entry: Mode Routing
進入時,先判斷走哪一層:
- 讀
references/dual-layer-guide.md - 解析用戶意圖 → Layer 1 或 Layer 2
- 不確定 → AskUserQuestion:「你要我直接幫你做這件事,還是要建一個可重複使用的 skill?」
- 鎖定 mode,中途不切換
Layer 1 Flow: Hands-on Mode
直接幫用戶自動化一個具體目標。
Phase 1: Requirements
問用戶(一次一題,用 AskUserQuestion):
- 目標:要自動化什麼?(網站 URL / API / CLI tool / 檔案格式)
- 操作:具體要做哪些操作?(生成、下載、上傳、轉換、查詢)
- 輸入/輸出:什麼進去、什麼出來?
- 已知資訊:已經知道什麼?(有帳號嗎、試過什麼、有文件嗎)
總結需求,確認理解正確。
STOP gate: 用戶確認需求。
Phase 2: Discovery Plan
- 讀
references/exploration-methodology.md - 根據目標類型(Browser / API / CLI / File / Service)選擇對應策略
- 建立 checklist — 列出所有要發現的元素/端點/操作
- 建立 discovery notes 文件(用 methodology 裡的 template)
呈現 plan 給用戶。
Phase 3: Environment
根據目標類型建立執行環境:
- Browser: Playwright setup, auth/session, headed browser
- API: HTTP client, API key/token, base URL
- CLI: 安裝工具, 確認版本, 測試基本用法
- File: sample files, parsing libraries
- Service: SDK install, credentials, sandbox
STOP gate: 環境可用、auth 可用。
Phase 4: Exploration
讀 references/exploration-methodology.md 的 Phase 4 Core Loop。
對 checklist 中的每一項:
1. 觀察當前狀態
2. 假設操作方式
3. 測試假設
4. 驗證結果
5. 成功 → 記錄 / 失敗 → 調整 → 重試
6. 下一項
關鍵規則:
- 一次只探索一個元素
- 每次操作前後都留證據(screenshot / log / response)
- 記錄到 discovery notes(不是直接寫程式碼)
- 不要猜 selector / endpoint / flag — 測試確認
What ships with it
2 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 First seen · 302 lines · 187 tokens per session scan A b9bb8c6620ec
tool-builder is a skill published in the GitHub repository fagemx/prismstack (2 stars, last pushed 4mo ago), licensed MIT. It adds 187 tokens to every session and 3,025 once invoked, about $0.0009 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
ebay-sold-listings-search
All process output to user (progress updates, process notifications) follows the user's language.
youtube-channel-business-email
YouTube channel business email and contact extractor: accepts a channel id (UCxxx), handle (@name), or URL; navigates the channel About view; extracts the business email from the description text plus full channel metadata (name, id, country, subscriber count, view count, video count, joined date, external links…
webcrawler-deep-crawl
Deep-crawl any website from start URLs, return per-page LLM-ready text/markdown/HTML plus metadata (title, description, author, language, canonical URL, OG) and in-scope outbound links. Use when user mentions deep crawl website, recursive crawl, crawl a whole site, scrape entire website, scrape docs site, scrape…
1688-product-detail
Extracts comprehensive wholesale product data from 1688.com product detail pages: title, tiered pricing, SKU variants with dimensions/weight, product images, seller info, shop scores, buyer protection, cross-border flags, product attributes, coupon/promotion data, and review stats. Use when user mentions 1688…
amazon-listing-competitor-analysis-skill
This skill helps users analyze Amazon competitor listings by ASIN and produce structured competitive intelligence plus strategic opportunity points for their own go-to-market. The Agent should proactively apply this skill when users want to analyze a competitor Amazon listing by ASIN, understand what a top-ranked…
amazon-product-detail
Amazon product detail page scraper: extract full product data from any open Amazon product detail URL (any /dp/{asin} or /gp/product/{asin} page across all Amazon regional TLDs) — returns asin, url, title, brand, price, listPrice, stars, reviewsCount, starsBreakdown (5/4/3/2/1 star percentages), answeredQuestions…