Context Preserver

Problem It Addresses

Agents in long-running tasks often accumulate intermediate conclusions, file changes, configuration state, and decisions. When a session is interrupted, the model is switched, or an operation goes wrong, it can be hard to return to a reliable working state. context-preserver makes that state explicit by saving it as snapshots instead of leaving important context only in chat history or temporary variables.

How It Works

The main workflow is built around the snapshot lifecycle:
- Automatic snapshots: create snapshots at useful moments as recovery baselines
- Manual snapshots: create tagged snapshots to mark distinct checkpoints
- Restore: return working context to a previous snapshot
- Manage: list, inspect, and delete snapshots
- Import/export: move snapshot data across tools, environments, or sessions

A typical usage pattern is to define the context boundary first, then create an automatic or manual snapshot. When rollback is needed, list snapshots, inspect the relevant one, and restore it. If a snapshot only captures part of the state, verify key files and variables after restore.

Boundaries and Notes

This skill is useful for managing an agent's working state, but a snapshot is not the same as a full system backup. The provided documentation does not specify storage location, permission model, encryption, or snapshot size limits, so sensitive data should be checked before enabling snapshots in production. Automatic snapshot behavior may need configuration to avoid creating too many checkpoints during high-frequency loops.