auto-resume

name	auto-resume
description	Recover active change and next action from artifacts. Use on fresh session with existing work.
metadata	{"stage":"resume"}

auto-resume

Session recovery. Rebuilds context from durable artifacts, not memory or guessing.

First action: run node .agent/.automaton/scripts/get-context.mjs from the project root.

Preamble

auto-resume rebuilds context from durable artifacts, not from the user's description or the agent's training data. It does not modify artifacts, advance the stage, or start new work. It loads canonical artifacts in dependency order (spec first, then design, then plan) and reports what it found, what was blocked, and what comes next.

Loading discipline: start with artifacts needed for the current stage. Read project files when understanding the codebase helps rebuild accurate context for the next action. Read .agent/.automaton/references/CONTEXT-BUDGET.md when wider reads threaten context pressure.

Quality Gate

Before producing the recovery summary:

• Trust durable artifacts over memory.
• Report stale pointers plainly.
• Recommend a next skill only when recovered state has incomplete or blocked work. For verified completion, report no next lifecycle skill.
• Read references/quality.md when the summary becomes narrative recap.

Do

Load State

Read .agent/.automaton/references/ARTIFACT-LIFECYCLE.md for recovery order, stale-pointer handling, and stage handoffs.

Halt and report when:

• .agent/ does not exist or current.json is missing.

Recommend auto-onboard and stop. Do not attempt recovery without a state file.

If work is complete or absent, read .agent/steering/ROADMAP.md only to surface pending phases as context.

Verify Artifact Integrity

Check that canonical_spec, canonical_design, and canonical_plan resolve when present. If any pointer is stale, report it plainly; recommend auto-onboard for missing steering, auto-frame for missing SPEC.md, or auto-plan for missing PLAN.md.

Load Artifacts

Treat current.json as the only source for active change, stage, and canonical artifact pointers. Load artifacts in dependency order and stop at the current stage; read references/artifact-order.md for the full stage table.

Reconcile Execution Ledger

When stage is execute or verify and the project is a git repo, read the execution ledger before summarizing: git log --oneline -15 and git status --porcelain. Per-slice commits (slice N: ..., slice N gap-fix: ...) mark verified slices; match them against PLAN.md slice evidence. A dirty tree on top of the last slice commit is in-flight work for the next slice, not noise: name the touched files. When commits and PLAN.md evidence disagree, trust the commits and report the mismatch. Also run git worktree list: a stray worktree is the fingerprint of an interrupted parallel dispatch. Report it; do not remove it.

Surface Review State

If current.json contains engineering_review, read the ## Review: Engineering section from the canonical plan and include it in the resume summary.

Recovery Summary

Produce a concise summary in this shape:

**Active change:** [name]
**Stage:** [frame|plan|execute|verify|verified|resume]
**Artifacts loaded:** [list]
**What was done:** [1-2 sentences]
**What was blocked:** [1-2 sentences, or "nothing"]
**What comes next:** [specific next action, or "none - change complete"]
**Execution ledger:** [last slice commit + in-flight files, "clean", or "n/a"]
**Review verdicts:** [engineering: X, or "none"]
**Missing state:** [list or "none"]
**Roadmap:** [N pending / M total, or "not tracked"]

The goal is orientation, not transcription.

Recommend Next Skill

Use references/recovery-scenarios.md for the full routing table. The invariant: recommend the next lifecycle skill only when recovered state is incomplete or blocked. For verified completion, report no next lifecycle skill; if ROADMAP.md has pending items, surface them as optional future work, not an automatic auto-office-hours handoff.

Output

• Resume summary (the template above, nothing more)
• Artifacts loaded
• Review verdicts (if present)
• .agent/.automaton/state/current.json is read-only for auto-resume; stale pointers are reported, not silently repaired
• Missing or conflicting state surfaces as a warning in the recovery summary.
• Orient and stop (utility skill): when recovered state is incomplete or blocked, emit Next: <skill>; none when the active change is verified complete. The user picks the direction.

Rules

• Prefer durable artifacts over memory.
• Do not restart discovery if the current artifacts are sufficient.
• Escalate contradictions instead of guessing.
• Load artifacts in dependency order: spec first, not plan first.
• If steering is scaffold-only, report it plainly and recommend auto-onboard.
• Do not turn a completed verified change into an automatic auto-office-hours handoff.