context-management

Project Sentinel is an open-source Claude Code mod for long autonomous runs that separately tracks work progress, context usage with automatic handoffs, and prompt-cache health/expiry. It adds autopilot fresh-context handoffs, cache protection, permissions/guardrails, resource limits, activity, profiles, and an optional companion, and can be installed via the plugin marketplace.

A Claude Code mod that replaces compaction with a structured handoff: a fork of the session writes the handoff from the prompt cache, and the conversation continues from exactly that handoff.

A Claude Code plugin that adds `/compact micro`, a manual compaction mode which locally elides Read tool results (including images and PDFs) and strips thinking blocks while keeping the rest of the transcript intact.

A status bar for the Claude desktop app. It draws above your prompt and shows session quality, context fill, cache countdown, and your 5-hour and weekly limits, with Clawd acting out what the session is doing.

lcm is shared memory infrastructure for coding agents, replacing sliding-window forgetting with a persistent runtime for humans and agents, with SQLite persistence, DAG summaries and cross-session memory.

A reminder mod in the Baselane mods catalog that prompts you to run /clear or /compact when context window usage passes 75 percent at turn end.

A Claude Code mod that hands a long session off to a fresh one before the context fills up. It replaces auto-compact.

A Claude Code mod that holds subagent fan-out, heavy-context prompts and retry loops before they spend your usage window. It reads your plan's usage, counts agents across every session, and asks in Claude Code's own dialog.

conPACT is a Claude Code mod that automatically compacts a session's context after a task finishes via an MCP tool, and prompts to compact idle sessions before the prompt cache expires. It ships as an in-process plugin hooks mod in Claude Code 2.1.284+, with a Stop hook fallback over Remote Control, and also supports Codex/ChatGPT Desktop. Author reports ~4.7% spend savings and lower peak context, MIT licensed.

A Claude Code mod that keeps an idle session's prompt cache warm with silent pings, then compacts large sessions once while the cache is still warm.

A Claude Code plugin that keeps notes you pin with `/pin-note` for the whole session and re-injects them verbatim into the model's context after each compaction and `/clear`, via a SessionStart hook, with per-session persistence in `$.store` and an off-by-default toggle.

The author nearly wiped a Kubernetes namespace after approving an agent’s `oc delete --all` without reading it, then built a suite of 11 Claude Code plugins using the new function hooks API. The main mod, guardrails, blocks destructive commands like mass kubectl/oc deletes, rm -rf, force-push, and secret reads, and tells Claude the user blocked it. Other mods cover activity monitoring, desktop notifications, context tracking, usage limits, prompt coaching, command discovery, and a mod manager. Nine of eleven never call a model, there is no telemetry, and the project is free and MIT licensed.