跳至正文
OpenAgentSkill
浏览已开放 · 目录记录日期 2026-11-21 · 自动更新和账户服务暂时关闭。
浏览器与自动化

stakeholders

Use when the goal depends on people who are neither on your team nor your opponents — councils, regulators, media, landowners, sponsors, rival organisers, the undecided public. Maps who holds power over the outcome, what each actually wants underneath their stated position, and where the movable middle is. Distinct from threat, which models opposition to degrade.

skyf0xx · 已有静态检查记录

安装与使用

运行命令前先阅读来源与依赖说明。复制命令不会自动安装或执行。

npx skills add skyf0xx/gambit --skill stakeholders

技能说明


name: stakeholders description: Use when the goal depends on people who are neither on your team nor your opponents — councils, regulators, media, landowners, sponsors, rival organisers, the undecided public. Maps who holds power over the outcome, what each actually wants underneath their stated position, and where the movable middle is. Distinct from threat, which models opposition to degrade. display: stakeholder-table

Skill: stakeholders

Trigger: The goal depends on people who aren't on your list and aren't your opposition. A council that grants or refuses. A regulator. A journalist. A landowner. A rival organiser. A sponsor. The large undecided middle whose indifference is currently the binding constraint.

Also use when threat returns "no identifiable adversary" but progress is still blocked — that usually means the obstacle is a stakeholder with different interests, not an opponent with hostile intent.

Purpose: Map the field of people who affect the outcome and can't simply be organised or defeated. The people key in GOAL.json tracks your own side's delivery. threat models opposition as something to degrade. This covers everyone else — which, in most real coordination goals, is where the outcome is actually decided.

The central discipline: separate position from interest. What someone says they want is a position. Why they want it is an interest. Positions conflict far more often than interests do, and almost every unlock lives in the gap between them.


Voice & Tone

Even-handed and specific. You are mapping people, not judging them — a stakeholder blocking the goal usually has a coherent reason, and finding it is the work.

Resist the pull toward two camps. The useful output is a gradient, not a friends/enemies split. Someone at "reluctantly neutral" who could be moved to "quietly supportive" is often worth more than consolidating people already on side.

Never write a stakeholder off as irrational. "Irrational" almost always means their interest hasn't been identified yet.


Execution Sequence

1. Load Context

Read GOAL.json — goal, success criteria, people key, systemsNotes and riskNotes keys if present. Note especially any criterion marked influence: whoever makes that decision is by definition a stakeholder, and often the most important one.

2. Enumerate the Field

List everyone who affects the outcome or is materially affected by it. Push past the obvious — the missed stakeholder is usually the one who only appears when they object.

Prompts to work through:

  • Who has to say yes? Permits, venues, funding, platform access.
  • Who can say no, or slow it down? Not the same list.
  • Who is affected but has no seat — residents, workers, users, neighbours?
  • Who benefits if this fails? Rivals, incumbents, anyone whose position it threatens.
  • Who is watching and will form a view — media, a regulator, a membership base?
  • Who is currently indifferent but wouldn't be if this got bigger?

3. Position vs. Interest

For each stakeholder that matters:

STAKEHOLDER: [name or role]
  Stated position: [what they say — publicly or to you]
  Underlying interest: [why. What they're protecting, needing, or afraid of]
  Confidence in that read: high | moderate | low — [if low: what would resolve it]
  Power over the outcome: high | medium | low
  Current stance: opposed | resistant | neutral | receptive | supportive
  Movable to: [realistic best stance] — via [what would actually move them]
  Cost of moving them: [time, concession, exposure, dependency created]

Where confidence in an interest is low, say so and offer bmad-deep-recon — acting on a guessed interest is how outreach backfires. Do not fabricate a motive to complete the table.

4. Power / Interest Grid

Place each stakeholder. Standard treatment, and it holds up:

                    LOW interest          HIGH interest
  HIGH power    [keep satisfied]        [manage closely]
                 ...                     ...
  LOW power     [monitor]               [keep informed]
                 ...                     ...

Two failure modes to call out explicitly when you see them:

  • Effort concentrated on high-interest / low-power people because they're pleasant to talk to and already agree. This feels productive and changes nothing.
  • A high-power / low-interest stakeholder ignored until their interest is triggered by the effort's own success — at which point they arrive late, uninformed, and hostile.

5. Find the Movable Middle

The largest available gain is usually not converting an opponent or consolidating an ally. It's moving several neutrals one step.

MOVABLE MIDDLE
  [stakeholder] — [current] → [achievable] — [the specific thing that moves them]

Rank by (power × distance movable) ÷ cost. Name the top one or two, not the whole list.

6. Coalition and Conflict Check

  • Coalition potential: which stakeholders share an underlying interest and don't know it? That shared interest is a coalition that doesn't exist yet.
  • Conflict risk: which stakeholders have interests that directly clash? You may not be able to satisfy both — if so, say which one the goal actually needs, and that choosing is a decision (decide), not a communications problem.
  • Association cost: does a stakeholder's support cost you standing with another? Name it. This is a live risk in any contested public effort.

7. Elicit

You are working from public information and inference. The user has met these people.

Before I write this down:

  - Which of these reads is wrong? You've actually talked to them.
  - Anyone important I've missed entirely?
  - Is there history between you and any of them I should know about?

History between parties usually outranks the analysis. Take the correction and rerun the affected rows rather than defending the grid.

8. Update GOAL.json

Replace the stakeholders array with the hig

阅读完整来源 ↗