create-boss
by vogtswcreate-boss turns boss chats, notes, emails, and project artifacts into a reusable skill for judgment, managing up, and persona modeling. Install it in Claude Code or OpenClaw to build real manager profiles or entrepreneur archetypes, then refine outputs with correction workflows and reusable Playbooks-ready boss guidance.
This skill scores 81/100, which means it is a solid directory listing candidate for users who want to turn boss communication patterns into reusable agent skills. The repository gives enough evidence for install decisions: clear triggers, two operating modes, install docs, parser/tooling, and concrete output structure. Its main limitation is that the core skill relies on multiple internal scripts and prompt files, so adoption is stronger for users comfortable with a somewhat involved workflow than for those expecting a minimal one-file skill.
- Strong triggerability: SKILL.md names explicit commands and natural-language triggers for creating, listing, rolling back, and deleting boss skills.
- Real operational substance: the repo includes parsers, prompt templates, archetypes, and a writer utility rather than only descriptive prose.
- Good install decision value: README and INSTALL.md explain supported data sources, installation paths, dependencies, expected outputs, and a validation loop.
- Core setup is not fully self-contained in SKILL.md; important install and dependency details live in README/INSTALL.md, including required Python packages.
- Repository guidance is partly bilingual and workflow-heavy, which may slow first-time users who want a simpler end-to-end example inside the main skill file.
Overview of create-boss skill
What the create-boss skill does
The create-boss skill turns messy boss evidence into a reusable operating model: how this leader judges work, how to manage upward with them, and how they typically speak under pressure. It supports two paths: build from real materials like chats, emails, meeting notes, and comments, or start from entrepreneur archetypes such as Elon Musk, Jeff Bezos, Steve Jobs, and Jensen Huang.
Who should install create-boss
This is best for people who need better manager-specific outputs, not generic leadership advice: ICs trying to communicate upward, staff/product/project leads preparing reviews, and Playbooks users who want a repeatable boss profile they can invoke later. If your real job is “help me predict what this boss will question and how to present work so it lands,” the create-boss skill is a strong fit.
Why use it instead of a normal prompt
A normal prompt can imitate tone once. The create-boss skill is more useful when you want a durable structure with separate judgment, management, and persona layers, plus versioning and correction workflows. That matters because the real adoption blocker is usually not generation, but refinement: “that’s not how they decide,” “they care about risk first,” or “they sound harsher in reviews than in 1:1s.”
How to Use create-boss skill
Install create-boss in Claude Code or OpenClaw
For a project-local Claude Code install, run from your repo root:
mkdir -p .claude/skills && git clone https://github.com/vogtsw/boss-skills.git .claude/skills/create-boss
For a global Claude Code install:
git clone https://github.com/vogtsw/boss-skills.git ~/.claude/skills/create-boss
For OpenClaw:
git clone https://github.com/vogtsw/boss-skills.git ~/.openclaw/workspace/skills/create-boss
Optional developer dependency:
pip3 install -r requirements.txt
If you only want to invoke the skill inside the agent, you usually do not need to run the Python tools manually.
What input create-boss needs to work well
The best create-boss usage starts with evidence that reveals decision patterns, not just tone. Prioritize:
- critiques and follow-up questions on projects
- edits to plans, weekly updates, or strategy docs
- comments during delays, risk spikes, or resource conflicts
- praise, rejection, urgency, and direction-change moments
Weak input: polished all-hands speeches or generic bios.
Strong input: raw comments showing what gets approved, rejected, escalated, or reframed.
A high-quality request looks like:
- boss name or archetype
- role/context
- your working relationship
- 5-20 representative examples
- what you want back: review simulator, upward-management guide, or persona clone
How to prompt create-boss for better outputs
Use trigger phrases like /create-boss, analyze my boss, or build a Musk-style boss. For create-boss for Playbooks, be explicit about the reusable artifact you want.
Better prompt:
“Create a boss skill for my VP of Product. Use these meeting notes, Slack excerpts, and doc comments. Focus on how she judges roadmap risk, what information she wants first, and how I should escalate bad news. Output clear judgment, management, and persona sections.”
Even better:
“Weight examples from launch delays and resource conflicts more heavily than casual chat. If patterns conflict, note the condition where behavior changes.”
That extra instruction improves the model because it separates stable rules from situational behavior.
Files to read first and suggested workflow
Read in this order:
SKILL.mdfor trigger phrases, modes, and workflowINSTALL.mdfor platform-specific setup and sample validation flowREADME_EN.mdorREADME.mdfor command examples and supported data sourcesprompts/for how intake, analysis, merging, and correction are structuredarchetypes/andbosses/to inspect output shape
Practical workflow:
- Start with one boss and a small evidence set.
- Generate the first draft.
- Correct obvious misses using concrete statements like “In roadmap reviews, he asks about owner and deadline before strategy.”
- Re-run or update until the generated skill predicts reactions reliably.
create-boss skill FAQ
Is create-boss good for real managers or only famous archetypes?
Both, but real-manager mode is the main value. Archetypes are faster for brainstorming or simulation, while real evidence produces more useful managing-up guidance. If you need actionability rather than inspiration, install create-boss for actual source material.
Is this beginner-friendly?
Yes, if you already have material to feed it. The hardest part is not setup but input quality. Beginners often overprovide generic conversation and underprovide review comments, escalation moments, and edited documents. You do not need deep prompt engineering, but you do need representative evidence.
When should I not use the create-boss skill?
Skip it if your manager is too new, you have almost no examples, or your need is generic “how to influence stakeholders” advice. It is also a poor fit if you want clinical personality analysis; this skill is optimized for work behavior, judgment patterns, and communication tactics.
What does create-boss add over ordinary prompting?
The create-boss guide value is structure and repeatability. The repo separates persona from decision logic and managing-up advice, includes correction handling, and supports listing, rollback, and deletion commands. That makes it easier to maintain over time than repeatedly re-explaining your boss in ad hoc prompts.
How to Improve create-boss skill
Feed patterns, not just volume
More text is not automatically better. For stronger create-boss usage, include moments where the boss made tradeoffs: speed vs quality, risk vs ambition, process vs output, autonomy vs escalation. A short set of sharp examples beats a huge dump of bland conversation.
Use concrete corrections after the first draft
The repo’s correction flow is one of the most valuable parts. Do not say “this feels off.” Say:
- “He does not start with strategy; he starts with timeline risk.”
- “She sounds calm in writing but sharp in live reviews.”
- “He tolerates uncertainty if ownership is clear.”
Concrete corrections help the skill update the right layer: judgment, management, or persona.
Improve create-boss for Playbooks workflows
If you use create-boss for Playbooks, tie the output to recurring situations: project reviews, status updates, resource asks, launch-risk escalations, and postmortems. Ask the skill to generate examples for each scenario. This makes the resulting boss profile easier to reuse inside operating playbooks instead of leaving it as a static character sketch.
Watch the main failure modes
Common misses:
- overfitting to tone instead of approval logic
- treating one bad day as a stable trait
- mixing public style with private decision rules
- producing flattering summaries instead of usable guidance
To improve create-boss skill results, iterate on conflicting examples, label contexts clearly, and keep asking: “What would this boss approve, reject, question first, and expect from me next?”