Skill: smart-resume

Description

Reads recent work context — STATUS.md, todo, open plans — to infer what's in progress and propose the single most useful next action. Executes on confirmation.

Trigger phrases:

• "smart resume"
• "pick up where we left off"
• "what were we working on"
• "what's next"
• "continue"
• "resume"

Steps

Step 1 — Read context (parallel reads)

Simultaneously read:

1. [MEM]/STATUS.md — session number + last change summary
2. [MEM]/todo.md — open items (High/Medium priority)
3. [MEM]/plans/ — glob for non-archived plan files; read the most recently modified

Step 2 — Check for in-flight work

If STATUS.md or todo references a specific file being edited, check its current state:

• Grep for // TODO or // IN PROGRESS markers in the mentioned file (use offset+limit — never read whole large files)

Step 3 — Infer the single most important next action

Rank candidates by specificity:

1. A plan file with Status: In Progress that has an identified next step
2. A High-priority todo item marked [ ]
3. The last change summary from STATUS.md — what's the natural follow-on?
4. A Medium-priority todo

Step 4 — Propose with full context

## Smart Resume — Session [N]

**Last session:** [STATUS.md summary]
**Open work:** [specific todo or plan step]

**Proposed next step:**
[Specific action — e.g., "Add the routing line for appXxx in routes.js around line 200"]

**Files involved:** [list]

Go? (yes / skip / show me all open items)

Step 5 — Execute on confirmation

If user says yes / go / do it:

• Follow plan-before-edit.md — show plan, wait for approval, then edit
• After edit: update STATUS.md + todo.md

If user says skip:

• Show next candidate from ranked list

If user says "show me all open items":

• Print the full todo.md High+Medium sections + all open plan files

Notes

• Always follow plan-before-edit.md before touching any code
• This skill reads context only in Steps 1-2 — no edits until Step 5 after explicit confirmation
• If no clear next step exists, say so: "No open work found — what should we work on?"