openlxcn/openlx-weixin-baimindan

OpenLX WeChat Official Account Publish Whitelist-Free Skill

No more repeatedly logging into the WeChat Official Account backend to modify the IP whitelist when changing computers or networks. It does not write articles, format content, or generate images; it is solely responsible for reliably delivering pre-prepared WeChat Official Account publishing requests through the OpenLX long-term online gateway. The first-time setup still requires scanning a QR code for authorization via the WeChat Open Platform, and all WeChat permissions, security, frequency, and content rules remain in effect.

Positive scenarios described in the instructions can trigger this skill automatically; what is explicitly authorized by the user is the gateway action, and there is no need to manually input the skill name each time.

Execution

1. Only run python3 scripts/manager.py check-update --agent <codex|claude|cursor> when using this skill. First run or every 7 days check; failure does not block the current task, no background service, no automatic updates.
2. Read HANDOFF_CONTRACT, lock the user-provided target official account, action, and original prepared data. Return PREPARED_PAYLOAD_MISSING or TARGET_ACCOUNT_MISSING if missing—do not supplement.
3. Credentials are read from the environment variable OPENLX_WEIXIN_API_KEY or the user's existing secure storage, never written to the skill or state files. Login, authorization, and rights reuse AUTHORIZATION.
4. Use scripts/gateway.py <handoff.json> to forward the prepared request as-is. CREATE_DRAFT requires payload.is_draft=true; public submission requires explicit publication confirmation from the user this time with explicit_publication_confirmation=true and payload.is_draft=false. Do not automatically select accounts, do not upgrade drafts.
5. After creating a draft, query and verify using the same media_id; timeout returns SUBMITTED_UNVERIFIED, do not blindly retry. A publish_id merely indicates submission; only report PUBLISHED after a real query returns success.

Updates require user confirmation; if there are local modifications, back them up first, then explicitly confirm the overwrite. See UPDATE. For diagnostics, run scripts/doctor.sh --agent <agent>. This skill is completely independent; it does not read or invoke any other skill, and does not depend on content factories or AG1–AG6.