task-flow-resume

Guides resumption of interrupted task work by verifying recorded progress against the repository tree.

1|Updated Jan 14, 2024
One-click install
npx skills add https://github.com/Eyhenij/rt-tools --skill task-flow-resume-eyhenij
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: task-flow-resume
Source: https://github.com/Eyhenij/rt-tools/tree/main/.claude/skills/task-flow-resume
Command: npx skills add https://github.com/Eyhenij/rt-tools --skill task-flow-resume-eyhenij

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? When returning to unfinished work in a new session, AI agents often re-ask the owner questions already answered, redo completed stages, or trust stale handover notes. This Skill defines how to resume work from recorded progress files instead of interrogating the owner or re-deriving decisions. ## Core Features & Use Cases - State verification against the tree: Checks the "Where we stand" section in progress.md against git status and branch history before doing any work, fixing divergences immediately. - Structured progress recording: Provides templates for rewriting the state line, logging decisions with reasons and rejected alternatives, and recording stage revisions without editing the plan. - Handover distrust protocol: Treats session handovers as unverified input, recounting all numbers on the current commit before reuse. - Use Case: A new session opens on branch TASK-142 where three of six stages are done. The Skill instructs the agent to read progress.md, verify the state with git commands, mark completed stages with command-output evidence, and continue with the next stage without asking the owner anything. ## Quick Start Resume the unfinished task on the current branch by reading the task folder's progress file and verifying its state line against the git tree before continuing work.

Frequently Asked Questions about task-flow-resume

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I resume unfinished work in a new AI coding session?

Read the task folder's progress.md state line first, then verify it against the tree using git status and git log before doing any work. Fix divergences in the record immediately, since the next session will trust the record over the tree.

How should an AI agent track progress across multiple sessions?

Rewrite a "Where we stand" section in progress.md rather than appending to it, recording state, current stage, completed work, next step, uncommitted changes, and PR status. Confirm each stage with command output, not a list made by reading.

Should I trust session handover notes when resuming work?

No. Handover notes lie outside the tree and describe the day they were written. Recount every number from the handover on the current commit and take the next step from the committed progress file, not the handover.

When should this resume pattern not be used?

Do not use it for starting new work; that is covered by the separate task-flow-start pattern. It also does not apply once a task is fully closed and handed off, which follows the task-flow-close pattern.

What should an agent do when a task is handed in for review?

The session does not end. The agent takes the next assigned task from the epic plan or work queue in the same turn, without offering the owner a choice that was already assigned at planning.