hope/skills/bond/SKILL.md
Assesses team fitness and composes agent teams. Use when "set up a team", "team for this", "should I use agents", "design a team", "how many agents", "agent team".
npx skillsauth add saadshahd/moo.md bondInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
Assess whether a task benefits from parallel agents. Most tasks are better solo. A team only pays off with genuine independence across modules.
Read the task scope. Evaluate internally:
Search the codebase to identify module boundaries and file ownership zones.
If any answer is no on file independence, fitness = solo. If scope is 3 or fewer story points, fitness = solo. Otherwise, fitness = team.
Do not output assessment reasoning in any form — no headers, no bullet points, no labeled sections. Go straight to Step 2.
One bold sentence stating your assessment, then AskUserQuestion with three options:
If user picks Solo or Subagents, state the approach in one sentence. Done.
If user picks Team, proceed to Step 3.
Identify 2-5 independent workstreams from the task. For each:
Run coupling check internally: for each role pair, "if A changes X, does B break?" Merge roles where yes.
If coupling remains after merging, proceed to Step 4 before presenting. Otherwise skip to Step 5.
One bold sentence naming the coupling. Then AskUserQuestion:
Repeat for each coupling pair. Then proceed to Step 5.
One bold sentence: "N roles, M workstreams, zero file overlap." Then present roles via AskUserQuestion (multiSelect) so the user reviews all roles at once:
MODEL: [opus/sonnet/haiku]
OWNS:
- path/to/files
TASK:
- what this role builds
DEPENDS ON:
- nothing / [specific handoff]
After user reviews, present a separate AskUserQuestion:
If plan mode is active: write the team spec to the plan file — role names, owned files, models, tasks, spawn order, and sequencing constraints. State that creation will happen after plan approval. Done.
If not in plan mode:
No output beyond confirmation that agents are running.
User approves before any team is created. Bond assesses and recommends; user decides mode, roles, and boundaries. Bond asks only bond questions — scope, requirements, and approach belong to intent and shape.
development
--- name: shape description: Resolves technical HOW decisions — architecture choices, technology selection, and design patterns — from a defined spec or intent. Distinct from hope:intent (which clarifies WHAT to build): shape starts when the goal is clear but the technical path is not. Use when: needing an implementation roadmap, choosing between architectural approaches, or resolving design trade-offs before coding. Triggers on: "shape this", "architecture for X", "how should I build", "system
development
Generate a project-level CLAUDE.md from stack detection and user-selected rule categories. Use when starting a new project, onboarding a repo, or when the user says "seed claude.md", "create project rules", "set up CLAUDE.md", "configure this project for me", or wants to establish coding conventions.
development
Turn rough ideas into clear work orders before planning or building. Use when request is vague like "add a button", "make it better", "fix the thing". Triggers on ambiguous or underspecified requests. Produces a brief with scope, acceptance criteria, and stop conditions.
testing
--- name: consult description: Simulates expert panels, compares documented positions across thought leaders, and synthesizes anonymous recommendations grouped by concern. Invoke when facing design tradeoffs, architecture decisions, repeated failure modes, or domain questions where multiple perspectives would reduce decision regret. Triggers on: expert names, "panel", "debate", "what would [X] say", "stuck on", style requests. --- Simulate expert perspectives by reasoning from documented positi