
Squad Explore
Explore an unclear implementation topic in your repo, get a direction report, and seed phased squad tasks—without the agent writing production code yet.
Install
npx skills add https://github.com/steloit/squad-skills --skill squad-exploreWhat is this skill?
- /squad-explore [topic] workflow: validate topic, clarify if needed, then deep codebase exploration
- Hard gate: does NOT write code—only direction report and phased squad tasks
- Clarification interview: one round, max 2 questions when scope or “why” is missing
- Mandates reading squad shared.md and principles.md before proceeding
- Self-sufficient topics (e.g. dark mode toggle) skip clarification and go straight to exploration
Adoption & trust: 1 installs on skills.sh; trending (+100% hot-view momentum).
Recommended Skills
Microsoft Foundrymicrosoft/azure-skills
Azure Aimicrosoft/azure-skills
Azure Hosted Copilot Sdkmicrosoft/azure-skills
Lark Eventlarksuite/cli
Running Claude Code Via Litellm Copilotxixu-me/skills
Setup Matt Pocock Skillsmattpocock/skills
Journey fit
Primary fit
Canonical shelf is Build / PM because the skill turns ambiguity into an implementation plan and squad board tasks after codebase exploration. PM subphase fits exploration-and-planning work that precedes direct frontend or backend edits.
SKILL.md
READMESKILL.md - Squad Explore
> Shared context: read `../squad/shared.md` for project config & auth, pipeline levels, status transitions, API endpoints, and error handling. > Safety principles: read `../squad/principles.md` — **mandatory, not optional.** ## `/squad-explore [topic]` — Explore & Plan **When to use**: You have a vague idea or problem but don't know *how* to implement it. This skill explores first, reports direction, then seeds the squad board with phased tasks. **This skill does NOT write code.** --- ### Procedure ``` ① Receive and validate topic If topic is missing (no argument): → Immediately enter the clarification interview (skip to ①-B). ①-A Check for missing context (NOT word count): A topic lacks context if ANY of these are true: - No indication of which part of the codebase is involved - The "why" is completely absent (what problem does this solve?) - The scope is unbounded ("improve everything", "refactor") If context is missing → ①-B If the topic is self-sufficient (e.g. "add dark mode toggle to settings page") → skip to ② ①-B Clarification (one round, max 2 questions via AskUserQuestion): - "What problem are you trying to solve or what outcome do you want?" - "Is there a specific area of the codebase you suspect is involved, or is it unknown?" Do NOT ask more than 2 questions in this round. ② Deep codebase exploration (Task → Explore agent) Launch a Task subagent with subagent_type="Explore". Pass the following prompt — fill in <TOPIC> and <PROJECT> before launching: ─────────────────────────────────────────────── You are performing a pre-implementation exploration for the topic: "<TOPIC>" Project: <PROJECT> Thoroughness: very thorough Investigate the following areas IN ORDER and report findings for each: A. PROJECT STRUCTURE - List top-level directories and their roles (1 line each) - Identify main entry files (main.ts, index.ts, app.ts, server.ts, etc.) - Read key config files: package.json (dependencies), tsconfig, vite.config or equivalent B. TOPIC-RELEVANT CODE - Find all files, modules, and components directly related to "<TOPIC>" - Identify existing patterns used for similar features (search by keyword) - Trace the data flow: where does data enter, how does it move, where does it exit? - Note any existing abstractions that could be extended vs. replaced C. PAIN POINTS & GAPS - Identify missing abstractions, obvious duplication, or inconsistent patterns - List all modules that "<TOPIC>" would need to touch - Identify potential conflicts with existing code or dependencies D. TECHNOLOGY CONSTRAINTS - Which libraries are already in use that are relevant? (from package.json) - What patterns does the framework enforce? (routing, state, DI, etc.) - What is the test/build/lint setup? Return your findings as a structured report with section headers A–D. For every claim, cite the exact file path and line number if possible. If you cannot find evidence for something, say "not found" — do not guess. ─────────────────────────────────────────────── ② ½ Architecture planning (Agent → Plan subagent) Save the Explore agent's output as $EXPLORE_FINDINGS. Launch a second Agent subagent with subagent_type="Plan". Pass the following prompt — fill in <TOPIC>, <PROJECT>, and <EXPLORE_FINDINGS>: ─────────────────────────────────────────────── You are performing architecture planning for the topic: "<TOPIC>" Project: <PROJECT> ## Codebase Findings (from Explore agent) <EXPLORE_FINDINGS> ## Your Task Based on the above codebase findings, pro