Decision Navigator skill review: branching questions

Decision Navigator is a Claude Code skill that teaches the model to guide users through branching questions instead of dispensing advice upfront. The skill, authored by sickn33 and hosted on agentskill.sh, targets sales, marketing, product, and HR contexts where users feel stuck or overwhelmed.

Metric	Value
agentskill.sh rating	0 (0 ratings)
Installs	0
Security score	100/100
Content quality score	92/100
Upstream GitHub stars	39,235
Last updated	2026-06-21

The zero-install, zero-rating status is the first red flag. Despite a strong content quality score (92/100), no one has installed or reviewed this skill in the marketplace. For a sales-focused skill, that suggests either insufficient visibility, unclear value prop, or genuine friction in adoption.

What Decision Navigator Claims to Do

The core instruction is simple: “collapse the problem space progressively” by asking one clarifying question at a time, offering 3-5 concrete paths, and repeating until the user reaches a “leaf” where concrete steps make sense. The SKILL.md frames this as “Navigate, don’t lecture.”

The philosophy is sound. Most users stuck on career changes, business launches, or decisions don’t need more options; they need the option space reduced. Claude’s default behavior often skips this step and jumps to advice. Decision Navigator aims to force iteration.

The provided SKILL.md is unusually detailed. It includes:

• A four-step process (acknowledge, ask one clarifying question, branch based on answer, deliver concrete steps at the leaf)
• Specific guidance on option label format (2-6 words, no colons or sub-explanations)
• Branching depth targets (typically 3-4 levels before actionable steps)
• A full example flow (user wants to start a business; Claude narrows down through two levels of questions)
• Tone guidelines (warm but efficient, short sentences, no jargon)

The example itself is solid. When the user says “I want to start a business but I have no idea where to start,” Decision Navigator’s first question asks about motivation, not a vague “tell me more.” The options are mutually exclusive and concrete (specific idea vs. freedom vs. more money vs. not sure yet).

Installation and Trigger Conditions

To use this skill in Claude Code, you add the SKILL.md to your local project at .claude/skills/sickn33/decision-navigator/SKILL.md. The agentskill.sh listing includes a content SHA (3a20cb7) and an installed timestamp, suggesting the skill can be pulled via the agentskill.sh CLI.

However, the SKILL.md provides no explicit trigger conditions. The “When to Use This Skill” section lists phrases (“I don’t know what to do”, “help me figure out”, “I feel lost about”) and scenarios (starting a business, changing careers, learning something new), but these are human-readable heuristics, not patterns Claude will automatically detect.

This is a critical gap. Without explicit trigger metadata or LLM-friendly patterns, Claude may not know when to activate Decision Navigator versus its default behavior. The skill could require manual invocation in the system prompt (e.g., “If the user seems stuck, use the Decision Navigator skill”) or depend on Claude Code’s auto-discovery mechanism, which is not documented here.

Declared Capabilities and Platforms

The marketplace record lists:

• Skill types: unknown
• Platforms: unknown
• Capabilities: not declared

This is vague. Based on the SKILL.md alone, the skill is a prompt pattern and decision-tree framework. It does not integrate with external APIs, databases, or tools; it’s pure conversation logic. This suggests it should work on any Claude model capable of maintaining multi-turn branching (Claude 3.5+), but the lack of explicit platform support means you may hit friction on Claude.ai (the web interface) if it doesn’t load agentskill.sh skills natively.

Comparison with Alternatives

Approach	Strengths	Weaknesses
Decision Navigator (sickn33)	Detailed branching guidance; respects user agency; designed for sales/HR soft skills	Zero installs; no platform declared; trigger conditions vague; untested at scale
Claude default (no skill)	Works everywhere; no friction; flexible	Often skips clarifying questions; dumps options/advice upfront; no structured branching depth control
Generic “structured interview” prompt	Easy to write inline; no dependency	Requires bespoke tuning per use case; no best practices baked in; not reusable across projects

Decision Navigator’s real advantage is systematizing something Claude doesn’t do by default: progressive reduction of the problem space with user consent at each step. It teaches the model to ask “what kind of problem is this?” before “here are 10 solutions.”

However, that advantage only materializes if the skill is actually invoked. With zero installs, there’s no evidence it solves a real pain point well enough for developers to reach for it.

Failure Modes

Several risks stand out:

Zero real-world validation. The skill has never been installed or rated. The SKILL.md is well-written, but prompt patterns often fail in practice due to edge cases, model drift, or context length constraints. A three-level branching question could easily consume 2k-3k tokens, leaving little room for follow-up or context from other skills.

Vague trigger conditions. Without explicit patterns Claude can detect, the skill relies on manual invocation or built-in auto-discovery (which agentskill.sh doesn’t document clearly). In practice, this likely means developers forget to use it.

Sales-focused but unproven. The target categories are sales, marketing, product, and HR. These are high-value domains where a good branching pattern could unlock real workflows. But zero installs suggests no one in those communities has found it yet, or it hasn’t been marketed well.

Broad applicability, but shallow implementation. The SKILL.md teaches how to branch, but it doesn’t solve downstream problems. Once you reach a leaf and give “concrete steps,” who enforces them? How does the skill track if the user actually acted? Decision Navigator is a conversation pattern, not a CRM or project management integration.

Content quality vs. install gap. The 92/100 content quality score is high, but the zero-install count shows that quality alone doesn’t drive adoption. The skill may be too abstract (“here’s how to ask questions better”) rather than solving a concrete pain point users recognize immediately.

When to Use This Skill

Genuinely useful cases:

• You’re building a chatbot that helps users navigate career decisions, business pivots, or similar high-uncertainty scenarios. The branching pattern keeps Claude from overwhelming the user.
• You want Claude to ask clarifying questions before you invoke downstream skills or tools. Decision Navigator can narrow the problem space so your subsequent skills have better context.
• You’re in sales or product and want a reusable pattern for discovery calls. The “extract, ask one question, branch” loop is portable.

Skip Decision Navigator if:

• Your use case is binary or pre-defined (e.g., “choose between Plan A and Plan B”). You don’t need three levels of branching.
• You’re building a system where speed matters more than user agency. The branching loop adds latency and tokens.
• Your problem already has explicit decision logic elsewhere (a decision tree in your database, a workflow engine, etc.). Decision Navigator duplicates that work.

Takeaways

Decision Navigator is a carefully written prompt pattern that systematizes something Claude doesn’t do by default: narrowing problem spaces through iterative, user-driven questions. The SKILL.md itself is one of the more thorough skill implementations on agentskill.sh, with clear examples and tone guidance. However, zero installs and zero ratings mean it has never been validated in production. The marketplace listing also lacks platform clarity and explicit trigger conditions, which could explain the adoption gap. For teams building decision-support chatbots in sales, HR, or product discovery, the pattern is worth reading; for immediate use, the lack of real-world evidence and rough integration edges make it a risky bet over a bespoke prompt or a more-established alternative.

Further reading

• Decision Navigator on agentskill.sh - Official skill listing with live installation option and user feedback mechanism.
• agentskill.sh documentation - The marketplace hub for Claude Code skills, including installation instructions and feedback API.
• Claude Code skills overview - Context on Claude’s local IDE and skill architecture (general reference).
• Prompt patterns in conversational AI - Academic context on how structured prompting improves dialogue quality.