Agent Exploration
Research multiple topics in parallel and synthesize findings into evidence-backed reports.
Installation
- Make sure Claude is on your device and in your terminal.
Skills load from
~/.claude/skills/when Claude Code starts up — so you need it on your machine first. If you don't have it yet, install it once with the command below, then runclaudein any terminal to verify.One-time setupnpm i -g @anthropic-ai/claude-codeAlready have it? Skip ahead.
- Paste into Claude Code or into your terminal.
This copies the whole skill folder into
~/.claude/skills/agent-exploration-compozy/— the SKILL.md plus any scripts, reference docs, or templates the skill ships with. Works for every skill.What this command does
- Makes a new, empty temporary folder.
- Downloads the skill's folder,
.agents/skills/agent-exploration, fromgithub.com/compozy/compozy(branchmain), the GitHub repository it's listed from. You get the version that is there now, not a copy we keep. - Copies that folder to
~/.claude/skills/agent-exploration-compozy/. - Deletes the temporary folder.
Installing runs none of the skill's files; it only downloads and copies them. When you later use the skill, Claude follows its SKILL.md and may run scripts that come with it. Read them first.
Faster alternative (instruction-only skills)
Skips the clone and grabs only the SKILL.md file. Don't use this if the skill ships Python scripts, reference markdowns, or asset templates — they won't be downloaded and the skill will fail when it tries to load them.
Quick install (SKILL.md only)Sign up to copy - Restart Claude Code.
Quit and reopen Claude Code (or any other agent that loads from
~/.claude/skills/). New skills are picked up on startup. - Just ask Claude.
Skills auto-activate when your request matches the skill's description — no slash command needed. Trigger phrases live in the skill's own frontmatter; you can read them in the “What this skill does” section above.
Prefer to read the source first? Open on GitHub.
When Claude uses it
Coordinate native explorers for explicitly requested multi-area research with scoped written results and an evidence-backed synthesis.
What this skill does
Agent Exploration
Use when explicitly requested for research across independent areas with written results. Use the current harness's native explorers and configured model/effort; an explicit user override wins. Do not invoke an external runtime or create an agent registry for this workflow.
Scope and dispatch
- Resolve the question and output path from the request or active task. Ask only for genuinely missing scope; creating the requested directory needs no additional confirmation. Default an unspecified artifact location to
.audits/exploration/<timestamp>/and state it. - Reuse current grounding. Scout only enough to identify independent questions; no tool-call or source-count quota. Choose as many useful slices as capacity and scope justify, up to 8;
--agentsis an upper bound, not a target to pad. - Read
references/dispatch-rules.mdandassets/explorer-prompt.mdonce when preparing dispatch. Send each explorer its question, source scope, exact output path, and the scoped-write contract once. Includeassets/analysis-template.mdwhen a structured artifact is needed; do not duplicate the contract in several embedded documents. - Create the output directory, then dispatch independent slices as capacity permits. Prompts need durable copies only for a long/resumable run. Every worker is read-only except for its named analysis file; no commits or unrelated edits.
- Inspect each completed artifact and its material citations. A failed slice does not invalidate successful siblings. Continue or correct the affected worker; do not launch a fresh whole round for a missing heading or confirmation sentence.
- Write
analysis/summary.mdwith the answer, evidence links, disagreements, remaining gaps, and actionable conclusions. Distinguish completed and unavailable slices; never claim an unperformed investigation succeeded.
Use analysis/NN_analysis_<slug>.md for each slice. Preserve externally required schemas; otherwise keep sections as short as their evidence permits. references/checklist.md is an inspection aid, not an extra approval stage. Routine single-file lookups need no delegation or research artifact.
Related skills
Co-Marketing Partnerships
coreyhaines31
Find ideal partners and plan joint marketing campaigns with other companies.
Research Toolkit
rrrrrredy
Write long-form research reports with fact-checked analysis in English or Chinese.
Idea Sprint
solanabr
Validate crypto startup ideas and decide whether they're worth building.
Jobs to be Done Canvas
product-on-purpose
Map customer motivations across functional, emotional, and social dimensions.