OKR lifecycle management skill for AI agents. Install it so your AI employees can create, track, update, and report on OKRs autonomously.

npx openskills install fractalmind-ai/okr-manager-skill

Then invoke in your agent:

npx openskills read okr-manager

Copy SKILL.md into your agent's skills directory (e.g., .agent/skills/okr-manager/SKILL.md).

• Create OKRs with enforced quality standards (measurable success criteria, dependency chains, concrete deliverables)
• Track progress by updating KR statuses with completion evidence
• Heartbeat audit — periodic check-ins that identify blocked KRs and push work forward
• Report status in a compact, scannable format

For AI employees and human+AI collaboration, the most reliable structure is:

• Objective = why this work matters
• KR = what observable result proves progress
• Task / Milestone = how execution is staged, assigned, and unblocked

In practice, that means:

• Keep KRs result-oriented — they should answer “did the world actually get better?”
• Keep Tasks/Milestones execution-oriented — they hold implementation steps, owners, and sequencing
• Do not let a task list masquerade as KRs

Good

• Objective: Make AI task routing stable and auditable
• KR1: 80% of tasks are assigned within 10 minutes with a verifiable receipt
• KR2: 90% of completed tasks report back to the source channel with PR / issue / evidence links
• Tasks: persist assignment state, verify agent receipts, post source-channel callbacks

Weak

• Objective: Improve AI collaboration efficiency
• KR1: Build agent manager
• KR2: Integrate Slack
• KR3: Improve UX

Those are implementation tasks, not key results.

The skill is configurable via your agent's workspace. Set these values as needed:

Every OKR created through this skill is validated against a checklist:

• Objective is outcome-oriented (verb + outcome, not activity)
• Success Criteria has a quantifiable metric
• Success Criteria is binary verifiable (done / not done)
• Every KR has a status marker
• Every KR describes an observable result, not just an implementation step
• KR dependencies are explicit
• Tasks / milestones are separated from KRs when execution detail is needed
• Every KR has a concrete deliverable + verification method
• Priority is marked (P0 / P1 / P2)
• Owner is assigned

Missing any item? The skill flags it before writing to the OKR file.

See examples/ for:

• okr-template-en.md — English OKR template
• okr-template-zh.md — Chinese OKR template
• heartbeat-integration.md — How to wire OKR audits into your heartbeat loop

AI agents (Claude Code, OpenAI Codex, custom agents) that manage projects with structured objectives. If your agent has a heartbeat loop and a markdown-based workspace, this skill fits right in.

MIT