explore
Technical exploration skill dispatcher. Exploration only — this skill never writes application code. Research notes, proposals, and repo comparisons are the only artifacts produced; the sole exception is a rare, clearly-labelled demo snippet inside a proposal, used to illustrate an approach and never shipped.
Inline Usage
Pass your intent directly on the invocation line — new proceeds immediately with no opening question when a description is provided.
/explore new how the auth middleware works
/explore new "caching strategy" --aspects
/explore diff https://github.com/org/other-repo
Critical Rules
- Exploration only — no code. This skill researches, compares, and proposes. It does not write application code, produce implementation plans or task breakdowns, or create Linear issues. The only permitted code is a rare demo snippet inside a proposal, clearly labelled as illustrative and not for production.
- Never execute workflow logic here — this file only parses args and dispatches
- Step 0 always runs first — no exceptions
- Unknown verb → run
help.md — never error silently
- Pass all remaining args through — workflow receives
$REMAINING_ARGS unchanged
- Markdown output: soft-wrap prose, never hard-wrap — when any explore workflow or agent writes a
.md artifact (proposals, research notes, summaries, comparisons), write each paragraph as one continuous line; do not insert manual newlines to wrap prose at a fixed column width. Newlines still separate paragraphs, list items, headings, and code fences.
- Model tiers, never model IDs — the
explore agents declare metadata: model-tier: light|standard|heavy in frontmatter and workflows use model-tier: tokens; the platform maps tiers to concrete models (see references/model-tiers.md). Never hardcode a provider model ID (such as claude-*) in this skill.
Step 0: Parse Arguments
The raw invocation args (filled by Claude Code / OpenCode slash commands): $ARGUMENTS. If this line is not filled in, read the verb and remaining args from the user's current message.
VERB="[first non-flag argument, or empty]"
REMAINING_ARGS="[everything after VERB, preserving order and flags]"
case "$VERB" in
"") VERB="help" ;;
"explore") VERB="new" ;; # alias: /explore explore → /explore new (pre-rename muscle memory)
esac
Step 1: Dispatch to Workflow
Read and execute references/workflows/{VERB}.md, passing $REMAINING_ARGS as the argument string.
If references/workflows/{VERB}.md does not exist, fall back to references/workflows/help.md and note the unknown verb.
Workflow Index
- new (
references/workflows/new.md) — research a technical problem, compare approaches, generate parallel proposals
- diff (
references/workflows/diff.md) — compare two repos for structural differences
- allow (
references/workflows/allow.md) — pre-approve permissions for background agents
- help (
references/workflows/help.md) — print command reference
Agent Index
- proposal-writer (
agents/proposal-writer.md) — writes technical proposals; used by new
- researcher (
agents/researcher.md) — researches technical approaches; used by new
1---2name: explore3description: Pure technical exploration: research a problem space, compare technical approaches, compare repos, and generate decision-oriented proposals. No code is written. Triggers on: "explore new", "explore diff", "explore allow", "explore help", "technical exploration", "research approaches", "explore options", "compare repos", "how does X work".4license: MIT5---67# explore89Technical exploration skill dispatcher. Exploration only — this skill never writes application code. Research notes, proposals, and repo comparisons are the only artifacts produced; the sole exception is a rare, clearly-labelled demo snippet inside a proposal, used to illustrate an approach and never shipped.1011## Inline Usage1213Pass your intent directly on the invocation line — `new` proceeds immediately with no opening question when a description is provided.1415```16/explore new how the auth middleware works17/explore new "caching strategy" --aspects18/explore diff https://github.com/org/other-repo19```2021## Critical Rules2223- **Exploration only — no code.** This skill researches, compares, and proposes. It does not write application code, produce implementation plans or task breakdowns, or create Linear issues. The only permitted code is a rare demo snippet inside a proposal, clearly labelled as illustrative and not for production.24- **Never execute workflow logic here** — this file only parses args and dispatches25- **Step 0 always runs first** — no exceptions26- **Unknown verb → run `help.md`** — never error silently27- **Pass all remaining args through** — workflow receives `$REMAINING_ARGS` unchanged28- **Markdown output: soft-wrap prose, never hard-wrap** — when any explore workflow or agent writes a `.md` artifact (proposals, research notes, summaries, comparisons), write each paragraph as one continuous line; do not insert manual newlines to wrap prose at a fixed column width. Newlines still separate paragraphs, list items, headings, and code fences.29- **Model tiers, never model IDs** — the `explore` agents declare `metadata: model-tier: light|standard|heavy` in frontmatter and workflows use `model-tier:` tokens; the platform maps tiers to concrete models (see `references/model-tiers.md`). Never hardcode a provider model ID (such as `claude-*`) in this skill.3031## Step 0: Parse Arguments3233The raw invocation args (filled by Claude Code / OpenCode slash commands): `$ARGUMENTS`. If this line is not filled in, read the verb and remaining args from the user's current message.3435```bash36VERB="[first non-flag argument, or empty]"37REMAINING_ARGS="[everything after VERB, preserving order and flags]"3839case "$VERB" in40 "") VERB="help" ;;41 "explore") VERB="new" ;; # alias: /explore explore → /explore new (pre-rename muscle memory)42esac43```4445## Step 1: Dispatch to Workflow4647Read and execute `references/workflows/{VERB}.md`, passing `$REMAINING_ARGS` as the argument string.4849If `references/workflows/{VERB}.md` does not exist, fall back to `references/workflows/help.md` and note the unknown verb.5051## Workflow Index5253- **new** (`references/workflows/new.md`) — research a technical problem, compare approaches, generate parallel proposals54- **diff** (`references/workflows/diff.md`) — compare two repos for structural differences55- **allow** (`references/workflows/allow.md`) — pre-approve permissions for background agents56- **help** (`references/workflows/help.md`) — print command reference5758## Agent Index5960- **proposal-writer** (`agents/proposal-writer.md`) — writes technical proposals; used by new61- **researcher** (`agents/researcher.md`) — researches technical approaches; used by new