Add esper:explore skill for pre-planning investigation

Context

The current esper workflow goes straight from idea to plan (/esper:plan or /esper:fix). There is no structured way to investigate feasibility, explore trade-offs, or brainstorm approaches before committing to a plan. The existing skills follow a consistent pattern: a skills/esper-<name>/SKILL.md file with frontmatter, steps, and tool usage instructions. There are currently 10 skill directories under skills/.

Exploration output will be saved to .esper/explorations/ — a new directory that does not yet exist.

Approach

1. Create skills/esper-explore/SKILL.md with the following workflow:

   • Step 1: Check setup — run esperkit config check, read .esper/CONSTITUTION.md and the current phase file for context
   • Step 2: Interview — use AskUserQuestion to understand what the user wants to explore (feature idea, architectural question, feasibility check). Ask whether they want codebase-aware exploration, open-ended brainstorming, or both
   • Step 3: Explore — for codebase-aware mode, use the Task tool with subagent_type: "Explore" to search the codebase for relevant patterns, files, and constraints. For brainstorm mode, discuss trade-offs, alternatives, and approaches conversationally
   • Step 4: Summarize findings — synthesize exploration into a structured markdown document
   • Step 5: Save — write the exploration doc to .esper/explorations/NNN-slug.md (use a simple incrementing ID based on existing files). Print the path and suggest next steps: /esper:plan to formalize, /esper:fix if it's a bug, or defer. Explorations are consumed during /esper:phase — the phase skill reads them, asks the user which ideas to include, and archives included explorations to .esper/explorations/archived/

2. Create lib/exploration.js — a module mirroring lib/plan.js for managing exploration files in .esper/explorations/. Commands:

   • list — list all explorations (exclude archived/), with optional --format json
   • next-id — scan existing explorations and return the next available zero-padded ID
   • archive <filename> — move an exploration to .esper/explorations/archived/
   • get <filename> — print frontmatter of a single exploration as JSON

3. Add exploration subcommand to bin/cli.js — route esperkit exploration <list|next-id|archive|get> to lib/exploration.js, following the same pattern as the plan subcommand.

4. Update skills/esper-phase/SKILL.md — add Step 3b to read explorations before the interview using esperkit exploration list, ask user which to include, and archive selected ones after phase is defined using esperkit exploration archive.

5. Update the explore skill's Step 5 to use esperkit exploration next-id for ID generation instead of manual file scanning.

Files to change

• skills/esper-explore/SKILL.md (create — the skill definition)
• lib/exploration.js (create — exploration management module: list, next-id, archive, get)
• bin/cli.js (modify — add exploration subcommand routing)
• skills/esper-phase/SKILL.md (modify — add Step 3b using esperkit exploration list and esperkit exploration archive)

Verification

• Run: node --test test/
• Expected: existing tests still pass, plus new tests for lib/exploration.js
• Edge cases: empty explorations dir, archiving already-archived explorations, ID generation with gaps

Progress

• Created skills/esper-explore/SKILL.md with 5-step workflow (check, interview, explore, summarize, save)
• Created lib/exploration.js with list, next-id, get, archive commands
• Added exploration subcommand to bin/cli.js
• Updated skills/esper-phase/SKILL.md Step 3b to use esperkit exploration list/archive
• Added 11 tests in test/exploration.test.js — all pass
• Modified: skills/esper-explore/SKILL.md, lib/exploration.js, bin/cli.js, skills/esper-phase/SKILL.md, test/exploration.test.js
• Verification: 64/64 tests pass