Language: English · 中文
YottaMeta's human-AI collaboration productivity skill: it converts “how to get things done with AI” into a repeatable collaboration protocol with a context brief, a plan-first gate, milestone delivery, verification, handover anchors and experience reuse.
Always-load at session start; a 30-second gate decides when the full protocol is needed, so simple questions stay simple.
Applies automatically to complex or long-running tasks, cross-session handovers, unreliable output, repeated rework or tasks with side effects.
No runtime, no daemon, no network calls: the skill is a protocol plus templates that any agent can follow on any platform.
It is the lowest common layer of cross-agent collaboration: any side may keep stricter local rules, and the stricter rule wins when they conflict.
People often fail with AI not because the AI is weak, but because the collaboration is sloppy: no context, no plan, no verification, no memory across sessions. Yuanban turns that around with a repeatable collaboration protocol:
- Context brief — background, goal, constraints, acceptance criteria.
- Plan-first gate — the AI proposes a plan; the user approves it before execution.
- Milestone delivery — one milestone at a time, each step visible and checkable.
- Verification — acceptance criteria, traceable evidence and user review, never blind trust.
- Handover and reuse — leave a session anchor, save lessons, make the next session smoother.
It is not a collection of motivational tips. It is a protocol with templates that can be copied and executed in any agent.
Yuanban is the lowest common layer of cross-agent collaboration protocols. It sets the minimum bar for “how to get things done with AI”, so any agent or team can keep its own stricter local rules (tighter state-file conventions, stricter release gates, higher evidence requirements). When they conflict, the stricter rule wins.
| Advantage | Description |
|---|---|
| Always-load, always light | Active from session start; a 30-second task gate prevents ceremony on simple questions |
| Executable, not inspirational | A fixed protocol unit: context brief, plan gate, milestones, verification, handover |
| Works across agents | Platform-neutral Markdown; no runtime, daemon or network required |
| Lowest common layer | Cross-agent minimum bar; any side keeps stricter local rules and the stricter one wins |
| Fixes the common failure modes | Missing context, direct action without approval, unverified output, lost session state |
| Verifiable, not theatrical | Acceptance criteria are checkable lists; unverified claims are labeled; evidence is real output |
| Self-correcting on exceptions | Detects common failure signals (acting without a plan, claiming completion without evidence, losing state) and stops to fix them with a one-line visible note |
| Focuses on the human | The user owns direction, judgment and final review; the AI handles execution and memory |
| Compounds over time | Lessons and effective practices are saved for the next collaboration (see yotta-learn) |
| Honest boundaries | Collaboration productivity only; no business, pricing or operations topics |
User: I need to migrate the legacy report pipeline to the new API.
AI: What is the background, goal, constraints and acceptance criteria?
User: Background: monthly report depends on deprecated endpoint. Goal: switch to new API.
Constraint: no downtime, page already has data. Acceptance: dry-run passes, one live run clean.
AI: Plan: 1) inventory endpoint usage, 2) build adapter, 3) dry-run, 4) live switch.
Files touched, verification steps, open questions. Shall I proceed?
User: Approve.
If the AI drifts — acts without a plan, claims completion without evidence, or loses state across sessions — it stops and corrects itself with a one-line visible note. Say "run the yotta-partner check" any time you want it to self-audit.
The detailed templates live in references/collaboration_protocol.md; exception and edge-case
playbooks are in references/exception_playbook.md; full worked examples in
references/walkthroughs.md; common mistakes and fixes are in references/faq.md.
Pick any of the four methods below; the order is the recommended priority. Skill files always come from npm (GitHub can be slow without a proxy; npm supports mirrors).
# Optional China mirror: npm config set registry https://registry.npmmirror.com
npx -y @yottameta/yotta-partner --agent <agent-name> # install to the agent's default user-level skills dir
npx -y @yottameta/yotta-partner --dir <your-skills-dir> # point to the skills dir itself (e.g. ~/.codex/skills)
--agent <name>installs to that agent's default user-level directory;--listshows each agent's default directory.--dir <path>installs to the given directory; for agents not in the preset list, point--dirat their skills directory.- If the mirror has not synced the new package (404): add
--registry=https://registry.npmjs.org/(a proxy may be needed in China), or wait for the mirror cache.
git clone https://github.com/YottaMeta/yotta-partner.git <your-skills-dir>/yotta-partner
On the GitHub repository YottaMeta/yotta-partner, click Code → Download ZIP, unzip it and put
the yotta-partner folder into the agent's skills directory.
bash install.sh --agent <name> # install to the agent's default user-level directory
bash install.sh --dir <path> # install to the given directory
bash install.sh --list # list agents -> default directories
Method 1 uses the npm registry (npmmirror / npmjs) and does not depend on GitHub; Methods 2/3 use GitHub and may fail without a proxy in China.
MIT © YottaMeta — see LICENSE.