Find the Top Three: Strategic Priority Triage for Agents

Where this fits

A user tells your agent: "I feel stuck. I want more freedom, but I also need income, and I keep going back and forth on whether to build a product or grow my audience — what should I actually focus on?" Most agents respond with either a generic life-coaching list or an endless stream of clarifying questions. This skill gives agents a disciplined decision framework for turning messy, self-contradictory context into three defensible priorities — while keeping the user in control and the reasoning transparent.

Why agents benefit

• Builds a corrigible user model: forms early provisional judgments from the first evidence, then explicitly revises them as new facts surface, so advice sharpens across sessions instead of repeating generic categories.
• Separates facts from interpretations: the workflow enforces a strict distinction between confirmed facts, the user's own explanations, and model hypotheses — the agent knows what it knows and what it's guessing.
• Detects proxy goals: "get 100k followers" or "buy a house" are treated as stand-ins for deeper wants (freedom, security, credibility) and decoded before any ranking happens.
• Knows when to stop: explicit convergence criteria prevent both over-interviewing and premature answers — the agent stops collecting information only when more questions can't change the ranking.
• Never defaults to templates: priorities are generated from user evidence, not a fixed "health, wealth, career" scaffold.

Practical scenarios

Career crossroads — A developer is torn between a promotion track, freelance income, and building a personal brand. The agent reads their actual constraints (savings runway, time budget, market signals) and helps rank which move creates the most optionality this quarter.

Creator strategy — A content creator juggles platform growth, product launches, and audience research. The skill separates loud metrics (vanity views) from real strategic leverage by examining what each goal actually buys.

Life-stage reset — Someone returning from burnout lists ten "should-dos." The agent applies the convergence test: which items survive contact with their actual energy, finances, and relationships? Most get cut.

Add it to your agent workflow

Install it with:

npx skills add yunshu0909/yunshu_skillshub --skill find-top-three

Then invoke it when the user asks for priority triage rather than simple todo sorting. A minimal usage prompt: "I'm overwhelmed by my goals for the next six months — help me find the most important three things." The skill will drive a multi-round exploration, build a minimal decision model, and only converge when evidence is sufficient.

Compare before adopting

Benchmark against generic prioritization prompts or fixed frameworks like the Eisenhower matrix. Look for: does the skill maintain a persistent user model across sessions, or does it start from zero each time? Is the evidence discipline visible (facts vs. hypotheses) in its output? How recently has the author updated the modeling framework? Alternatives like custom chain-of-thought prompts are cheaper but lack the structured output templates and convergence logic this skill ships with.

Why it is worth tracking

With 752 stars and an actively maintained repository, this skill addresses a known failure mode in agentic advice: giving confident answers from thin evidence. The corrigible modeling, proxy-goal detection, and explicit convergence criteria are the kind of design decisions that matter when agents move from tactical to strategic guidance. Evaluate it whenever your agents need to give priority recommendations, and watch for updates to the modeling framework — that's where the real depth lives. See the skill page for reference.