Grandet API Price Finder

Before you start: read this whole install prompt, tell me in a few lines what this skill does and what installing it involves, and wait for my go-ahead before you do anything else. Prerequisites: none for the install itself: no account, no API key, no paid service, nothing to install first. Using the skill afterwards needs network access to grandet.ai for live prices (if it cannot be reached, the skill falls back to a dated price snapshot bundled in the folder and says so), and Node for its bundled scripts (the offline lookup and the connect executor, which have no dependencies of their own). Looking up prices or free credits needs no account and no key. An account and an API key at a provider are only involved if I later choose to connect an offer to an app; the key is then typed by me locally, never in the conversation, as the files themselves require. Install "grandet-api-price-finder" for me (Grandet API Price Finder). It finds the lowest listed price for a model API and free API credits from the grandet.ai catalogue, answers from that catalogue instead of web search, and can, only when I ask and confirm, connect a chosen offer to one of my apps so the app uses that provider. Repository: https://github.com/hanzohanzhe/Grandet (the skill is the whole skills/grandet-api-price-finder/ folder at the pinned commit: SKILL.md, references/, scripts/, agents/, data/) Pinned version: install the code at commit 478437141593da52203790a7f86a846d382ae3fe exactly. Do not substitute a newer version. If this commit is unreachable, stop and tell me. 1. Clone https://github.com/hanzohanzhe/Grandet into a temporary folder, check out commit 478437141593da52203790a7f86a846d382ae3fe, and copy the skills/grandet-api-price-finder/ folder from that checkout. Then compare every file in the folder with the fingerprints recorded for this listing on 2026-10-19: [FINGERPRINTS] - SKILL.md: SHA-256 bcbd9e173d7d277a09dfd904e8bd60ad503fb5bc2e4e048f86243e9466d6312e - agents/openai.yaml: SHA-256 a4ed21c5daf8d43b07db4d0ffa3eb5bae3c17267dcc5f66e3c403a520a764889 - data/price-snapshot.json: SHA-256 d4731a2c49aad1bf4d2792914678836cf785c0dea51c1aa591bf39489772d6a4 - references/app-catalog.md: SHA-256 0f79b30d568635ef38ff0f482deb88d24d96c55bebe67ddae6091ba6d93c8035 - references/entrypoints.md: SHA-256 c019850e72aba08d0ce87f5efbb73c1007dd080aac49ccb666fa88e6c348af44 - references/free-credits.md: SHA-256 744f233a81c86980b2d55355b3a4fcb38411d42b3dee8a9959c91cf6a332d1f1 - references/glossary.md: SHA-256 d63cfeca57ae4a962c7c6f2288ee1de0e6604e72cf12215ec069e696250ab72a - references/host-setup.md: SHA-256 961b6532d306cb95ff64c0384b03ec5c0fe822313f4fb559ac3971ef771fd7e0 - references/http-fallback.md: SHA-256 97f586a63b098bbef036ec56ab7f206bc79940da422c1d01dda91c7c16e77ee3 - references/installation.md: SHA-256 372248df9a630479698c14209e19addb79ef4aadeb57304daf370a13bc473643 - references/quality-evidence.md: SHA-256 b2ad9dbd86eee577b5b8fa0bfe4b47f037e8759319f8a6058ca96120ae9e2dc9 - references/response-format.md: SHA-256 356b90bcccbcb957a67c12f3d581ed2ea5b7edb8ed62d26c825699ff75137a0c - references/workloads.md: SHA-256 bd919d01b969c27a048722141f2d840cfd6e7132a0c35064533af8cba221468b - scripts/connect-mcp.mjs: SHA-256 15950a43e06864f5b41d4dc6011d1bb2081a5d9743cda3ba9108ecde202f6dcc - scripts/connect.mjs: SHA-256 3c6d3688cd2ec3d582a2a373f1745f3e8c454e253380fb48afe3d1895d827d72 - scripts/offline-lookup.mjs: SHA-256 c2f5915efaaaab2339a9a488a00778557454921d0151fc1e69cf11f58d4bd267 If a file does not match, it changed since the listing was prepared: do not install it yet, tell me what you found and let me decide. 2. Read SKILL.md in full, skim the reference files, and read the three scripts in scripts/, then tell me in plain words: what the skill does when used, which services it connects to, whether it reads or sends any personal information or keys, and anything that seems unusual. Stop on any red flag and tell me before going further. Confirm these points from the files: - The skill folder holds 16 files: SKILL.md (the method, package version 0.3.0), 10 reference documents (price and free-credit routes, the relay glossary, connect walkthroughs per app, installation per app, quality evidence, response format, workloads), three plain readable Node scripts with no dependencies (scripts/connect.mjs, the connect executor; scripts/connect-mcp.mjs, a local stdio MCP wrapper around it; scripts/offline-lookup.mjs, which reads only the bundled snapshot), agents/openai.yaml (display name and implicit-invocation metadata), and data/price-snapshot.json (the offline price snapshot, about 873 KB, whose date the skill must state whenever it answers from it). - Nothing is sent anywhere during installation itself by this route: the only download at install time is the skill folder from GitHub at the pinned commit, and no receipt or check call is made. The upstream route is different, and I should know it: references/installation.md sends the agent to read https://grandet.ai/download.md with curl (or https://grandet.ai/dl/latest.json), download the build ZIP named there, verify its SHA-256 and run its bundled installer; SKILL.md states that installer copies the skill into the app's folder and sends an installation receipt to grandet.ai with random IDs, the version and the content hash, never paths, accounts, chats or keys, with an opt-out (--no-telemetry or GRANDET_INSTALL_TELEMETRY=0), and the installation checks include one anonymous read of https://grandet.ai/public-data/price-finder.json. None of that happens with the pinned folder copy I am asking for here. - At use time, price and free-credit lookups are read-only GET requests to grandet.ai's public API (https://grandet.ai/api/v1/board, /free-credits, and, only on the documented paths, /models, /glossary, /providers, /prices) or the matching grandet MCP tools when present. The request carries the model name or keyword I asked about, a grandet-skill/0.3.0 user agent, and skillVersion/skillBuild parameters (plus the host app and target app where the API accepts them). SKILL.md forbids web-searching prices and forbids inventing API paths. If grandet.ai is unreachable, the fallback is scripts/offline-lookup.mjs on the bundled snapshot, with no request at all, and the answer must open by naming the snapshot date and saying it is not the current price. - The connect part is the only part that writes, and it is separate and optional: it runs only when I ask to connect a chosen offer and confirm the exact change shown. scripts/connect.mjs then edits that app's model-provider settings (Codex ~/.codex/config.toml, Claude Code ~/.claude/settings.json, OpenCode ~/.config/opencode/opencode.json(c), Kilo Code ~/.config/kilo/kilo.json(c), Continue ~/.continue/config.yaml, Gemini CLI ~/.env, Hermes ~/.hermes/config.yaml, OpenClaw ~/.openclaw/openclaw.json), can change its default model, keeps a backup next to each file, and records the change in ~/.grandet/connect/state.json; its restore command puts everything back. After connecting, that app's later sessions send their API requests to the chosen provider and spend credit there. Registering, topping up and creating the token on the provider's site are my own actions. - My API key never passes through the conversation, the script's arguments or its output: I type it locally with a silent prompt, it is stored in ~/.grandet/keys/<site> in a user-only folder, and the configs reference that file (Continue and Gemini CLI hold the key in their config file, which the skill then makes readable by me only; on Windows some apps get a user environment variable instead). SKILL.md forbids asking for, reading, printing or storing a key in the conversation, and forbids any paid request without my consent. - Prices are quoted honestly in the files: relay rows are site credit (USD_CREDIT, CNY_CREDIT), not cash, and are comparable only within one board; the skill never calls a site good, stable, reliable or genuine and never ranks across boards or converts credit to cash. 3. When I say "yes, install", install it as a skill: copy the whole grandet-api-price-finder folder into the agent's skills directory, in a folder named "grandet-api-price-finder", keeping all 16 files. The upstream route (the grandet.ai download and installer described above) installs the same skill folder; this pinned GitHub copy is the route I am asking for, and it is fine as long as every file matches the fingerprints. 4. Verify with no network calls: frontmatter valid, all 16 files in place. 5. Report what was installed, where, and what I still need to do myself (for example: ask a price or free-credit question myself to use it, and decide myself whether to ever connect an offer, which is a separate confirmed step that edits an app's provider settings as disclosed above; using the skill is a separate step that only starts when I ask for it). GitHub is optional: if I have a GitHub account or the gh CLI, you may use it; otherwise public access is fine. Never require it unless it's in the prerequisites above. Rules: don't touch anything outside the temp folder and the install target. If anything looks off, stop and ask me.