yunsii
UserA local-first control plane for AI coding agents, built on WezTerm, tmux, and git worktrees. / 本地优先的 AI Agent 控制平面，基于 WezTerm、tmux 和 git worktrees。
Categories
Indexed Skills (15)
task-ledger
Ledger development tasks to Feishu Base via dev-task-ledger.sh. Write roots from ~/.openclaw/tasks-allowlist.json (per agent). Use on accept/confirm/close of coding/ops work; not for pure Q&A.
adversarial-review
Multi-role adversarial code review (platform skill). Agent loads this skill and runs the co-located runner (find → refute → sandbox repro); humans only state intent (审一下 / 对抗审查). Works for OpenClaw Main and host TUI agents via user-level skill links. Use when user asks for adversarial review, acceptance recommends review, or runtime code changed before merge.
brainstorm
Multi-persona divergent→convergent brainstorm (platform skill). Agent loads this skill and runs the co-located runner (diverge → challenge → converge); humans only state intent (头脑风暴 / 帮我发散一下 / brainstorm / 想点子). Sibling of adversarial-review; reuses its provider layer. Use when the user wants ideas / options / approaches for an open-ended problem — NOT for reviewing existing code (that is adversarial-review).
yuns-engineer
Engineer judgment for concrete problems: evidence → options → decision → verify → mine transferable practice. Routes real capabilities when needed and distills design decisions + personal preferences. Load for adjudication, multi-option recommendations, or end-of-task experience capture. Not idea divergence (brainstorm) or code-diff review (adversarial-review).
dev-task
Allowlisted development (wezdeck (+ optional team roots in local config)) under OpenClaw: wezdeck defaults to primary master; claw worktrees when parallel/isolation needed; 团队仓 still prefers claw-*; human/Claw rails (H1/H2, C1/C2/C3), ACP access layer.
chrome-devtools
Drive the machine Chrome debug browser via OpenClaw-managed MCP (chrome-devtools). Use for page navigation, snapshots, screenshots, clicks, forms, console/network inspection — not for host shell risk.
exec-risk
Host shell MUST go through claw-run (gate then exec). Three layers: rules → Grok re-check → human only if still danger. Prefer claw-run over bare exec. Also: exec hygiene — short shell, scripts on disk via claw-script-run (no fragile python -c / node -e), split chains, gateway self-update only via detached unit.
wezterm-runtime-sync
Synchronize this repository's WezTerm runtime. Reuse a valid cached target from repo-root `.sync-target` or `WEZTERM_SYNC_TARGET` when available; otherwise list candidate homes, confirm one with the user, and run the skill-owned sync script.
error-closed-loop
Full error closed-loop for Dex (Main): diagnose, safe self-fix, verify, report; platform Exec failed handling; escalation with options. Load when handling failures or writing failure reports.
session-bridge
Host ↔ Claw Session Adapter Kit: list/capture host tmux panes, list/tail OpenClaw sessions, poke agent turns, lease-gated host-send-keys, bot-send, optional say-as-me, take/watch poller for focused-pane handoff. Use for Feishu/host interop — not a second TUI.
cross-repo-delegate
Cross-repo work tickets (platform skill). Use when the user wants to file a ticket from one project to another (提单 / 跨仓委托 / 开给别的仓库的单), or in the receiving project to list inbox, claim, challenge assumptions, reply, or close (有没有我的单 / 认领 / 关单). Agent loads this skill and runs the co-located runner; humans only state intent. Target repos come from the local allowlist config, not from this description.
human-run
Mandatory handoff when a human (not the agent) must run a script/command. Agent loads this skill, ensure-env, propose.sh --cwd (default --wait until x finishes), then the agent continues follow-up itself. Handoff script bodies must be non-blocking kickoffs (trigger/async ack) — never wait for CI/deploy inside x. Never paste multi-line scripts into chat. Agent self-exec is out of scope. Use when you would otherwise ask the user to paste/run a script, or the command needs TTY / host secrets / out-of-sandbox privileges.
worktree-recycle
Reset a long-lived linked dev-* workstation onto origin/HEAD after the round is delivered: soft preflight (beyond git), call worktree-task recycle, then project-specific post-init. Use when the user says 重置开发分支 / 重置工作站 / recycle this worktree / 收尾换下一轮, optionally with a next-task brief.
habit-weekly
Produce a stable personal development-habit weekly report from WezDeck observability: pane-scoped agent concurrency, Claude/Grok/Codex skills+CLI+MCP, CDP verify→iterate, hotkey intensity (Alt+l etc.), optional WakaTime summaries, and optional push to a private habit archive repo. Use when the user says 开发习惯周报 / 习惯周报 / habit weekly / 这周热键和 agent 习惯 / 出一份习惯数据报告. Prefer this over ad-hoc chat analysis once habit-report exists.
repo-hygiene
Repo hygiene audit and pre-commit gate for wezdeck: broken relative markdown links, bash -n, mermaid on staged docs, secret heuristics, line budgets. Use when installing commit hooks, running a full-repo size/link audit, or when the user asks for 卫生审计 / doc-link check / line-budget review. Not adversarial-review and not design-review.
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.