github-aw-memory

You can persist GitHub agent state across runs by applying cache-memory for ephemeral tracking, repo-memory for long-lived Git commits, or comment-memory for notes attached to issues and PRs. This ensures workflows avoid repeating previously completed tasks.

Prevent GitHub agentic workflows from repeating work by implementing deduplication using stable identifiers. Store these identifiers in cache-memory or repo-memory so subsequent runs can safely check the persistent workflow state before executing tasks.

Yes, you can use cache-memory for incremental processing in GitHub workflows. It provides ephemeral persistence ideal for short-lived progress tracking and metric baselines that expire safely, allowing incremental tasks to resume without duplicating previous steps.

To persist workflow notes directly on a GitHub issue or PR, use the comment-memory strategy. This approach scopes the context to the triggering source, keeping the persistent state closely tied to the specific issue or pull request.

Use repo-memory instead of cache-memory when workflow state must survive cache eviction and require auditable Git commits. Repo-memory provides long-lived persistence for GitHub agentic workflows, whereas cache-memory is designed for short-lived, ephemeral data.

Cache keys with colon timestamps cause Windows artifact failures because colons are restricted characters in Windows file paths. To prevent this, enforce filename and key timestamp formats that avoid unsafe characters when persisting GitHub agent state.