---
name: dealing-with-a-centralizing-boss
description: |
  Use when the user is working for, advising, evaluating, or deciding how to engage with an organization
  where authority radiates from a single person at the center — lieutenants report directly to them,
  jurisdictions overlap by design, real power does not live in any institution.
  Triggers on phrases like "my boss runs everything through himself", "this company is a one-man show",
  "the CEO won't delegate", "I'm stuck advising a strongman", "how do I survive under this kind of leader",
  "how do I influence a leader who trusts no one", "should I take this role under X".
  Do not use for: conventional hierarchical orgs where delegation is real, early-stage founder-led startups
  where centralization reflects genuine one-person bandwidth (the pattern looks similar, the prescription
  differs), or purely interpersonal bad-boss situations without the structural features below.
---

# Dealing with a Centralizing Boss

> *What this skill is about, in one sentence:*
> How to recognize when an organization is built around one person as the sole source of authority —
> and, once you know you're inside that structure, how to protect yourself, exercise influence,
> and avoid the traps that consume people who don't see the pattern.

## Where this comes from

This skill distills the "hub-and-spokes model of leadership" chronicled in Chapter 1 of Jeffrey Sonnenfeld
and Steven Tian's *Trump's Ten Commandments* (2025), with additional material from Chapters 3 and 7.
Sonnenfeld derives the pattern from three earlier bodies of work:

- **Alfred Chandler** (*Strategy and Structure*, 1962) and **Larry Greiner** (*Evolution and Revolution as
  Organizations Grow*, 1972) — early-stage entrepreneur leadership, where a single founder as the hub is
  normal and often functional.
- **Sonnenfeld's own *The Hero's Farewell*** (1988) — his typology of "Monarch" and "General" CEOs who
  cultivate intramural rivalries to remain irreplaceable.
- **Walter Bagehot** (*The English Constitution*, 1867) — "no man can argue on his knees", on how
  subservient posture manufactures conformity.

The book is not a primary source (it is Sonnenfeld's interpretation of Trump), so the operators below
are presented as **diagnostic patterns of centralizing leaders**, not as a reading of any single person.

## North Star

> *If Claude did not have this skill, would it usually give worse decisions when a user is navigating a
> centralizing org?*

This skill earns its place only when it changes what a user **does**: whether they take the job, how
they advise someone inside, when they try to influence the leader, how they protect themselves from
being disposable. Decoration about leadership philosophy — drop it. Keep only what shifts action.

## Opening Gate: is this actually a hub-and-spokes org?

Before applying anything below, run this check. Centralization that looks like hub-and-spokes may
actually be something else:

- **True hub-and-spokes** — jurisdictions overlap by design, lieutenants lack independent credentials or
  constituencies, acting/interim appointees dominate, the leader publicly plays deputies against each
  other, there is no succession plan. Apply this skill.
- **Early-stage founder-led** — a small company where the founder is still the only person with full
  context. Structurally similar, but driven by bandwidth not by control. The operators below partially
  apply; the counter-moves about preserving independent power bases are less urgent because the leader
  is not actively suppressing them.
- **Ordinary autocratic manager** — a bad boss in a healthy broader org. The org's own institutions
  (HR, board, legal) still constrain them. Do not use this skill; use ordinary bad-boss advice.

Also check three cross-cutting meta-gates that run through the entire *Ten Commandments* framework:

- **Method behind apparent chaos** (Sonnenfeld's core claim, Introduction): when a leader's behavior
  looks chaotic or impulsive, check whether it fits a small, repeated playbook. If it does, treat it as
  designed, not erratic.
- **Law of the instrument** (Abraham Kaplan, *The Conduct of Inquiry*, 1964): dominance-seeking
  leaders over-apply a narrow set of tools. Expect the next move from their known repertoire, not from
  the situation.
- **Public vs. private persona mismatch** (Chapter 5): the public show and the private encounter can
  be very different. Do not predict one from the other.

## The Operators

### 1. `hub-and-spokes-detection`

**Plain English:** The tell-tale organizational shape of a centralizer: everyone reports to the top,
jurisdictions are deliberately fuzzy, and no lieutenant has a real independent base.

**Source:** Sonnenfeld & Tian, Ch. 1. Underlying concept from Chandler (1962) and Greiner (1972).

**Detect:**
- Lieutenants hold multiple overlapping roles ("cabinet-level and cabinet-adjacent jobs") that exceed
  their expertise.
- Acting/interim appointees outnumber confirmed ones.
- Direct reports all report to the leader; no real second-in-command.
- Lieutenants chosen for dependence rather than credentials — professional networks, technical depth,
  or public stature would make them threats.
- The leader publicly pits deputies against each other.
- Cash flows through related-party entities (shell LLCs, hotel bookings, licensing fees) rather than
  transparent accounting — a sign the position is being treated as personal property (Ch. 4 material
  folded in here).

**Intent behind it:** Prevent any lieutenant from accumulating enough independent authority to box the
leader in or succeed them.

**Counter-move:**
- If you're evaluating a job: assume you will be interchangeable regardless of what you're told.
- If you're inside: maintain one form of legitimacy that does not depend on the leader's favor
  (external expertise, credentialed network, outside constituency).
- If you're advising from outside: your access is leverage only until it isn't; don't stake your own
  reputation on access that can be withdrawn without notice.

**Do not use when:** the org is genuinely early-stage and the founder is the hub by capacity
constraint, not by design.

### 2. `acting-vs-confirmed-leverage`

**Plain English:** Acting/interim appointees are maximally dependent on the leader; Senate-confirmed
(or equivalently board-ratified) appointees gain autonomy. A centralizing leader will prefer interim
status indefinitely.

**Source:** Sonnenfeld & Tian, Ch. 1.

**Detect:** Persistent preference for "acting" titles over formal confirmation; nominations withdrawn,
delayed, or never submitted; interim appointees kept dangling with the promise of a formal role.

**Intent behind it:** Interim appointees must keep pleasing the leader to secure the "real" job. Once
confirmed, the leader's leverage drops to firing or freezing out.

**Counter-move:**
- If offered an interim/acting role: treat it as probationary indefinitely.
- If a leader offers you a formal seat with real power, don't assume the offer is straightforward —
  confirm what it takes to remove you, and whether your confirmation is actually being pursued.

**Do not use when:** interim status is genuinely bureaucratic (awaiting scheduled confirmation
timelines in a functioning process), not a posture.

### 3. `plausible-deniability-via-vague-suggestion`

**Plain English:** The leader makes suggestive, ambiguous comments — not direct orders — about what
they want. Subordinates infer intent and act. When the action misfires, the leader disowns it.

**Source:** Sonnenfeld & Tian, Ch. 1, citing the "will no one rid me of this meddlesome priest" story
about Henry II of England (1170, the Thomas Becket assassination).

**Detect:**
- Directives arrive as musings, rhetorical questions, or sideways remarks rather than written orders.
- Subordinates are praised for "taking initiative" on things the leader later denies knowing about.
- When things go wrong, the leader does not claim the action; the subordinate is cut loose (examples
  in the book: Giuliani's unpaid legal bills, Navarro's jail time without executive privilege invoked).

**Intent behind it:** Offload risk and reputational damage onto the subordinate while keeping the
upside of the action.

**Counter-move:**
- **Never execute on inferred intent without written direction.** If the ask is consequential, ask in
  writing: "To confirm, you'd like me to…?" and get a written answer.
- If the leader won't put it in writing, treat the non-answer as the answer: the risk is yours alone.
- Keep your own contemporaneous record of what was asked and how.

**Do not use when:** the leader's culture is casual but written-record-friendly, and ambiguity is
laziness rather than strategy.

### 4. `elimination-tournament-selection`

**Plain English:** Candidates for a role are pitted publicly against each other in an extended
contest. The drawn-out tournament is the point — it brands all contenders as dependent on the
leader's favor, regardless of who wins.

**Source:** Sonnenfeld & Tian, Ch. 1 and Ch. 7. Compared in the book to the format of *The Apprentice*.

**Detect:** Public auditioning; leaks of opposition research between candidates; candidates
kept waiting in hotel rooms or antechambers; the leader publicly entertains replacing the near-winner
with a dark horse at the last moment.

**Intent behind it:** The process itself demonstrates that every contender's standing comes from the
leader. The winner emerges humbled, not empowered.

**Counter-move:**
- If you're a candidate: understand that participating in the tournament is itself a concession. The
  winner's position will be weaker than a normal appointment.
- If you can credibly opt out, opting out preserves your independent stature — and paradoxically
  sometimes increases your leverage to be recruited outside the process.

**Do not use when:** the selection is a genuine merit-based search and public comparison is for
transparency, not for demonstration of dependency.

### 5. `reach-before-public-commitment`

**Plain English:** A centralizing leader is far more open to disagreement in private, at the formative
stage of a decision, than in public afterward. Once they have announced a position, reversing it
requires overwhelming external pressure because public reversal reads as weakness.

**Source:** Sonnenfeld & Tian, Ch. 1 ("Those who want to influence him through constructive engagement
are most successful when they reach him at an early, private formative time") and Ch. 5.

**Detect:** Not a detection operator — a timing rule.

**Intent behind it:** Private disagreement does not cost the leader face; public disagreement does.

**Counter-move:**
- Influence window is **before** the public announcement. After the announcement, direct persuasion
  fails; only external pressure large enough to offset the face-saving cost works.
- Frame your input as giving the leader new information or alternatives they hadn't considered, not
  as correcting them. Never lecture, never moralize, never corner them in front of others
  (cross-reference Ch. 5 and the Zelenskyy Oval Office case).

**Do not use when:** the leader has a track record of genuine public reversals; some do exist, and
this rule is specifically about leaders whose face-saving cost is unusually high.

### 6. `loyalty-is-one-way-and-vengeance-is-selective`

**Plain English:** A centralizing leader expects total loyalty but does not reciprocate. Ex-insiders
who leave quietly are left alone; ex-insiders who go public face disproportionate retaliation.

**Source:** Sonnenfeld & Tian, Ch. 1, Ch. 7. Book cases: Pence ("Hang Mike Pence"), Sessions primary
loss, Bolton home raid, Krebs investigation — all after public criticism.

**Detect:**
- Long pattern of discarded ex-insiders with nothing to show for their service.
- Public attacks on former allies correlate with the ex-ally's public criticism, not with any other
  misstep.
- Cross-reference with operator 8 (`winner-vs-loser-by-constituency` in skill #8): retaliation
  targets tend to be people the leader perceives as having lost their sympathetic constituency.

**Intent behind it:** Enforce silence from current and future insiders by making the cost of public
criticism visible.

**Counter-move:**
- If you're planning to exit: the line that converts you from forgotten to targeted is **going
  public**, not leaving. Many ex-insiders live peacefully. Those who publish, testify, or criticize
  on television are the ones attacked.
- If you must go public, pre-commit to the full course — don't half-criticize. Half-criticism
  provokes retaliation without the moral standing of full criticism.
- Preserve your own independent constituency before exit; a former ally with real public standing is
  a harder target than one without.

**Do not use when:** you were never really "inside." The retaliation pattern applies to insiders who
break ranks, not to outside critics, who often receive milder treatment.

### 7. `invert-hierarchy`

**Plain English:** A centralizing leader elevates lower-ranked, less-credentialed people over their
nominal superiors. This strips independent authority from the institution and re-roots it in the
leader's personal favor.

**Source:** Sonnenfeld & Tian, Ch. 3 (military examples: Mattis/Hegseth, Milley/Caine), Ch. 1.

**Detect:**
- Junior officers, obscure figures, or outsiders-from-the-field placed above senior, credentialed
  ones.
- Retired or lower-ranked persons given authority over incumbents who outrank them.
- Attacks on the institution's criteria for legitimacy (the Federalist Society example in Ch. 3, once
  it stopped producing rulings the leader wanted).

**Intent behind it:** Dismantle institutional sources of independent legitimacy so that all authority
flows back to the leader.

**Counter-move:**
- If you're in the displaced senior role: your formal rank will not protect you. Your leverage is
  outside the institution — external networks, public standing, independent expertise.
- If you're watching from outside (advising a board, writing policy): treat the inversion as a signal
  that institutional constraints will be overridden, not respected, and plan accordingly.

**Do not use when:** an unconventional promotion is genuinely merit-based or a targeted reform, not
a pattern of systematic inversion.

### 8. `no-true-friends-only-followers`

**Plain English:** A centralizing leader's relationships are transactional, not affectional. They
may be personally warm, charming, even funny in private — but there are no real friendships in their
orbit, only people with current utility.

**Source:** Sonnenfeld & Tian, Ch. 5, Ch. 7. Book observation: the leader has no wolfpack of lifelong
friends, no ex-girlfriends with insightful things to say, no college reunion crew.

**Detect:** Absence of long-standing non-transactional relationships; everyone in the orbit is either
a follower or a current-utility partner.

**Intent behind it:** Not a deliberate tactic — a structural feature. Warmth and charm are real in
the moment but do not produce durable obligation.

**Counter-move:**
- Do not mistake warmth for friendship. The private charm described in Ch. 5 is compatible with
  instant discard when utility ends.
- Calibrate expectations: gratitude for past help is not a reliable currency. Every interaction must
  re-prove your current usefulness.
- If you value the relationship, engineer ongoing reasons for the leader to need something from you.

**Do not use when:** the leader shows a track record of durable commitments across years to people
who are no longer useful — a centralizer who passes this test is a genuinely different type.

## Cross-cutting pattern

Operators 1–4 describe how the structure is built and maintained. Operators 5–8 describe how to
operate inside it. The pattern is fractal: the same hub-and-spokes dynamic shows up whether the
"org" is a company, a political party, a negotiation table, or a family. The counter-moves scale
accordingly.

## Final-answer structure

When Claude uses this skill to advise a user, the response must end with:

### Judgment
- Is this actually a hub-and-spokes org? Which signals support that?
- Which operators are most load-bearing for this specific situation?

### What Would Change My Mind
- What would make me reclassify this as ordinary founder-led centralization?
- What would make me drop a particular counter-move (e.g., the leader has in fact reciprocated loyalty
  across time)?

### Next Action
- Concrete: do X, do not do Y, before Z.

## References

- `references/source-notes.md` — book chapter mapping and academic lineage for each operator
- `references/rejected-candidates.md` — candidates considered and cut
- `references/cases.md` — longer-form cases cited inline (Mattis/Hegseth transition, Pence Jan. 6,
  Giuliani unpaid bills, Navarro jail time, Scaramucci 11 days)