# Resume From Binnacle

> User-invoked shortcut to resume a multi-session run from its binnacle: locates .claude/binnacle.md at the current worktree root, delegates to the binnacle skill's resume procedure in full, optionally warms context with read-only explorers when the latest entry points at code it does not explain, and continues the work. Invoked manually as /resume-from-binnacle at the start of a cold session, to avoid retyping the resume prompt every time.

- Skill: `diegopherlt/resume-from-binnacle` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add diegopherlt/resume-from-binnacle`
- Raw SKILL.md: https://api.skillmd.com/api/skills/diegopherlt/resume-from-binnacle/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: DieGopherLT (https://skillmd.com/u/diegopherlt)
- Updated: 2026-09-10
- Page: https://skillmd.com/skills/diegopherlt/resume-from-binnacle

---


# Resume from binnacle

You were invoked at the start of a cold session to pick up a multi-session run. This skill adds
nothing to the resume contract — it is the deterministic entry point that saves the user from
retyping the same prompt. The procedure itself belongs to the `binnacle` skill; never reimplement
it here.

## Steps

1. **Locate.** Check for `.claude/binnacle.md` at the root of the current worktree. If it does
   not exist, stop with a clear message: this session is not inside a run's worktree. Do not
   search other worktrees, and do not create a binnacle — opening is task-planning's job.
2. **Delegate.** Invoke the `binnacle` skill and execute every numbered step of its *Resuming*
   section in order, without summarizing, reordering, or skipping any. Do not restate those
   steps here — that section is their single source of truth.
3. **Warm context, only if needed.** After the binnacle skill's plan read, check the latest entry's
   **Siguiente** and **Referencias**: if they point at code this session must understand and
   neither the entry nor the plan explains it, launch read-only Explore agents in parallel — one
   per distinct area — and fold their conclusions into your working context. The decision comes
   after the plan read on purpose: the plan often documents exactly what an explorer would go
   find. A well-written entry needs zero explorers; the default is none, and launching them for
   code already explained wastes the budget the binnacle exists to protect.
4. **Continue.** Start working the first pending task of the session in the main context, under
   the binnacle's *Convenciones de ejecución*.