diff --git a/README.md b/README.md index 3b1e86b..4a90cee 100644 --- a/README.md +++ b/README.md @@ -25,6 +25,7 @@ shapes instead of pointer chains. - `shop-agent` - Browser-driven Amazon/retailer shopping: search, compare, price-check, add to cart, and walk checkout — always stopping for human approval before purchase. Pairs with `web-use` for product/price data and browser context. - `shrimp` - Internal pass-through helper for `/shrimp` sub-agent dispatch. - `telegram-ui` - Telegram rich chat UI patterns: inline buttons, polls, edits, replies, reactions, media, and pins via OpenClaw. +- `tweetclaw` - Approval-gated TweetClaw workflows for X/Twitter search, source review, media, monitors, webhooks, giveaway draws, and posting. - `web-use` - Front-door routing for web search, fetch, extraction, protected backends, and real browser context. ## Companion kits diff --git a/skills/tweetclaw/SKILL.md b/skills/tweetclaw/SKILL.md new file mode 100644 index 0000000..5fb0847 --- /dev/null +++ b/skills/tweetclaw/SKILL.md @@ -0,0 +1,101 @@ +--- +name: tweetclaw +description: Use TweetClaw from OpenClaw for X/Twitter search, reply context, media, monitors, webhooks, giveaway draws, and approval-gated posting through the official @xquik/tweetclaw plugin. +--- + +# TweetClaw + +Use this skill when an OpenClaw task needs structured X/Twitter data or a +reviewed social-account action through TweetClaw. + +TweetClaw exposes 2 OpenClaw tools: + +- `explore` - local catalog search for supported endpoints and parameters. +- `tweetclaw` - optional live endpoint invoker for catalog-listed calls. + +## Install + +Install the official package: + +```bash +openclaw plugins install npm:@xquik/tweetclaw +``` + +Enable the optional live tool only when the workspace should allow TweetClaw +actions: + +```bash +openclaw config set tools.alsoAllow '["explore", "tweetclaw"]' +``` + +Verify runtime registration: + +```bash +openclaw plugins inspect tweetclaw --runtime --json +openclaw skills info tweetclaw +``` + +Agent-skill installers can install the packaged skill without enabling the +runtime plugin: + +```bash +npx skills add xquik-dev/tweetclaw +``` + +## Configuration + +For account-backed automation, store the API key in OpenClaw config from an +environment variable: + +```bash +openclaw config set plugins.entries.tweetclaw.config.apiKey "$XQUIK_API_KEY" +``` + +For read-only pay-per-use workflows, configure the local signing key: + +```bash +npm i mppx viem +openclaw config set plugins.entries.tweetclaw.config.tempoSigningKey "$MPP_SIGNING_KEY" +``` + +Do not paste API keys, signing keys, cookies, or account credentials into chat, +docs, logs, screenshots, issue bodies, or tool arguments. + +## Workflow + +1. Start with `explore` to find the endpoint, required parameters, and access + mode. +2. Use read endpoints first to gather source evidence, such as tweet search, + tweet replies, user lookup, follower export, media context, trends, or draw + inputs. +3. Draft social copy separately from execution. +4. Before any post, reply, delete, follow, direct message, monitor, webhook, + extraction, or account-scoped action, show the structured request and get + explicit user approval. +5. Use `tweetclaw` only for the approved endpoint and arguments. +6. After writes, report the returned identifier or status and any follow-up + verification path. + +## Side-Effect Boundaries + +- Never turn TweetClaw into unattended publishing. +- Never approve future social-account actions by implication. +- Never expose credentials or raw account session material. +- Do not invent endpoints. Use `explore` and the runtime inspection output. +- Keep user review between drafting, scheduling, posting, deleting, following, + direct messaging, and recurring automation setup. + +## Good Fits + +- Search tweets before drafting a response. +- Check tweet replies before running a giveaway draw. +- Export followers for review. +- Collect user lookup context. +- Upload media for an approved post. +- Monitor tweets after the user approves the monitor definition. +- Deliver webhook-backed follow-up automation with an explicit reviewed target. + +## Source Links + +- TweetClaw repo: +- npm package: