Read and write session state in WORK.md and FLOW.md with git checkpoints.
npx skills add https://github.com/nullhack/temple8 --skill run-session
Please help me install this Agent Skill. Skill: run-session Source: https://github.com/nullhack/temple8/tree/main/.opencode/skills/run-session Command: npx skills add https://github.com/nullhack/temple8 --skill run-session
Automates the lifecycle of multi-step sessions by ensuring state is consistently read at the start and written at the end, enabling agents to resume where they left off.
Start a session by initializing from existing WORK.md and FLOW.md, detecting the current state, and loading the active feature to resume progress.
High-intent search queries and answers about installing and using this skill.
Automate session state management by reading WORK.md and FLOW.md at the start of a session and writing updated checkpoints at the finish. This detects the current filesystem and git state, enabling AI agents to resume multi-step workflows exactly where they left off.
The best way to checkpoint an AI-assisted workflow is to use automated session detection from your filesystem and git history. This approach reads your active feature state and updates checkpoints with a traceable commit history, ensuring no context or decisions are lost when pausing.
Git-based state detection works by reading your existing WORK.md and FLOW.md files to auto-detect the actual session state from the filesystem and repository. It then updates checkpoints with traceable commit history, allowing controlled transitions and accurate resume points.
No specific dependencies or external components are required to enable session checkpointing. The approach relies entirely on reading and writing to your existing WORK.md and FLOW.md files, using git and filesystem state detection to manage transitions and resume capabilities.
Yes, you can use this session automation approach for team-based feature development. It ensures state is consistently read at the start and written at the end of a session, allowing teams to pause a feature in progress and resume later without losing context or decisions.
Multi-step AI agent workflows lose context between steps when session state is not consistently read at the start and written at the end. Automating this lifecycle with checkpointing and state detection prevents context loss and ensures controlled transitions across steps.
nullhackDesign and operate state machine workflows with FLOW.md and WORK.md.
lancekrogersAutomate festival task discovery, progress tracking, and traceable commits.
spacholski1225Execute implementation plans in controlled batches with human review checkpoints.
26mitch26Capture Claude session contexts including Git state, decisions, and pending tasks.
