| name | bili-sc-bridge-ops |
| description | Safely operate, diagnose, and repair this repository's Bilibili SC/gift/fleet bridge, including QR login, authenticated multi-room listeners, rendering, QQ desktop delivery, owner-only Bot control, persisted routing state, and recovery. |
Bilibili SC Bridge Operations
Locate the repository by sc_bridge/, Makefile, .env.example, and README.md; on the owner's Mac it is normally /Users/longmozhou/Documents/Codex/2026-08-04/wo.
Non-negotiable constraints
- Use Makefile commands backed by
venv/bin/python. Never use uv run. Do not delete or recreate venv/ unless dependency repair is explicitly requested; the known hidden-venv issue can break imports.
- Never reveal, quote, copy elsewhere, or commit
.env, QQ secrets/tokens, Bilibili cookies, or clipboard contents. Credential diagnosis may expose only key presence, value length, duplicate definitions, and whitespace metadata.
- Prefer
make bili-login and the ignored runtime/bilibili-auth.json over manually copying Cookie. Treat that file and its refresh_token as secrets; status checks may expose only validity, UID, and nickname.
- Inspect
git status before edits and preserve unrelated changes. Do not include uv.lock, _preview.py, or package-lock.json in a commit unless the owner explicitly includes them.
- Start with read-only evidence. A real QQ send, GUI automation, process termination, credential regeneration, service installation, or external push requires explicit authority for that action.
- Preserve owner-only controls. Guild @/DM uses
QQ_OWNER_USER_ID; QQ friend C2C uses the distinct QQ_OWNER_C2C_OPENID. Never infer that one ID proves the other.
- Never run WorkBuddy and this bridge against the same QQ Bot Gateway. Use separate bots or a different WorkBuddy messaging channel.
Routing
- For modes, login lifecycle, configuration precedence, commands, status semantics, storage, and verification, read references/operations.md.
- For an actual incident, read only the relevant section of references/troubleshooting.md.
- Treat the repository
README.md and current source/tests as authoritative when they conflict with this Skill; update this Skill after changing behavior.
Minimum completion evidence
After code changes run make test, git diff --check, and focused compilation or tests proportional to the change. Distinguish simulated/unit evidence from a live QQ/Bilibili test. Restart a running process before claiming it uses newly edited code.
