sk-explore

First-link discovery. Gather repo context and write a structured report that feeds into sk-alternatives or sk-planning.

SKILL.md

GitHub

Explore

When to use

  • Starting work when the problem, scope, or repo context is unclear.
  • The user asks to explore, research, or understand X in this repo.
  • Before sk-alternatives or sk-planning when context is missing.

Goal

Produce a self-contained 0 - EXPLORE.md so the next phase can proceed without re-exploring.

Slug and flow folder

  1. Derive a short kebab-case slug from the goal (confirm with the user if not provided).
  2. Create .agents/flows/sk-<slug>/ if needed.
  3. Create or update RUNBOOK.md with all phases pending; mark 0 as in-progress.
  4. Write 0 - EXPLORE.md and set row 0 to done with a one-line summary.
  5. Add .agents/flows/sk-*/ to .gitignore if this is the first flow in the repo.

Output

0 - EXPLORE.md includes:

  • Goal: restated scope.
  • Files and Context: relevant files and inline facts.
  • Findings: what exists, what's missing, patterns, constraints, edge cases.
  • Assumptions and Open Questions: unknowns with acceptance impact.
  • Risks: known blockers or high-risk areas.
  • Next Step Recommendation: sk-alternatives or sk-planning with a one-line reason.

Steps

  1. Clarify — restate goal and scope. If ambiguity is high, ask or list assumptions.
  2. Search — find relevant code, tests, docs, and config.
  3. Read — inline the relevant snippets or facts.
  4. Synthesize — summarize current state, constraints, and gaps.
  5. Recommend — sk-alternatives if the approach is unclear, sk-planning if it is clear.
  6. Write — save the report and update RUNBOOK.md.

Do not edit non-report files.