task-pickup

name	task-pickup
description	Resume an existing task cleanly by recovering current state, selecting the next bounded step, and preparing a clear handoff for the following run.

Skill: Task Pickup

Purpose

Resume an existing tracked task cleanly by recovering current state, selecting the next bounded step, and preparing a clear handoff for the following run.

When To Use

• continuing work under projects/[name]/work/[task-slug]/
• delegated implementation sessions
• resuming work after interruption or context loss

Recommended Roles

• Builder
• Historian
• Oracle, when the task is review-focused

Required Inputs

• TASK.md
• HANDOFF.md
• CONTEXT.md
• relevant project files
• optional NOTES.md if the task flow adopts it
• existing files under runs/ when recent run evidence matters

Steps

1. Read TASK.md to confirm the task scope and success criteria.
2. Read HANDOFF.md for current operational state.
3. Read CONTEXT.md for stable task facts.
4. Read only the project files needed for the next likely step.
5. Identify the highest-priority unfinished in-scope action.
6. Complete one bounded slice rather than partially touching multiple areas.
7. Before ending, update HANDOFF.md with current state, what changed, and the next action.
8. If the workspace uses task-browser metadata, align the task's metadata with the real state using framework/tools/task-browser/metadata-cli.mjs; for example set the picked-up task to active, blocked, or review as appropriate. In framework versions with task-browser action history, pass real provenance such as --role Builder, --session-tool pi, or --session-id ... when useful and available; leave provenance unset rather than inventing it.
9. Move any newly discovered stable facts into CONTEXT.md or project library files as appropriate.
10. Write a run log.

Outputs

• one completed bounded task slice or a clearly documented blocker
• updated task handoff
• updated stable task facts when needed
• task-browser metadata aligned when the workspace uses task-browser
• a run log entry

Stop Conditions

• the current slice is complete or a real blocker is documented
• the next run can continue from files without guessing

Pitfalls / Anti-Patterns

• skipping HANDOFF.md
• letting HANDOFF.md become historical buildup instead of current operational state
• putting temporary status in CONTEXT.md
• changing scope without updating task state
• leaving task-browser metadata stale in workspaces that use the task-browser board
• leaving partial work without recording what remains
• reading too much irrelevant project code before choosing a step

Related Files / Tools

• framework/TASKS.md
• projects/[name]/work/[task-slug]/TASK.md
• projects/[name]/work/[task-slug]/HANDOFF.md
• projects/[name]/work/[task-slug]/CONTEXT.md
• projects/[name]/work/[task-slug]/runs/
• framework/tools/task-browser/metadata-cli.mjs when task-browser metadata is used