context-management

A Reddit submission introducing "Context Diet", a Claude Code mod intended to prevent large tool outputs from flooding the context window. The post text only contains submission metadata (author, subreddit, link/comments placeholders) with no repository link, installation steps or demo media, so the mod cannot be verified from the source alone.

Keeps a long-running agent's working notes through every compaction: briefs the summarizer, re-attaches the notes after the summary, and shows them in a panel you can steer.

When context usage reaches a threshold you set, write a handover note first, then compact automatically, and hand the note to whoever continues: this conversation after compaction, and the next session in the same project.

A Claude Code plugin that trims long Bash/PowerShell output before Claude reads it, keeping the head, tail and error lines, and saving the full output to a file Claude can open without a permission prompt.

Context-driven development for Claude Code: setup, spec, plan, implement, review, archive, handoff, and revert.

A Claude Code plugin that lazy-loads glossary definitions into the model's context when your prompt mentions matching terms.

A Claude Code function-hooks plugin. When the main conversation has been idle for idleMinutes (default 50 minutes), it compacts before the prompt cache expires and passes AGENTS.md's Compact Instructions to every compact. It skips running turns, conversations below minMessageTokens, and cases beyond cutoffMinutes, leaving the reason in the transcript.

Claude Code Mod: continuously shows context/rate-limit usage on the status line, summarizes the work near a limit, writes handoff.md and commits it, then asks whether to continue at the next startup.

A Claude Code function-hook mod that shows a one-line gauge above the prompt for context fill, cost and hot rate limits, toasts at 60/80/90%, preserves your instructions and edited files through compaction, and adds a /ctx context breakdown pane.

Hands a long session off cleanly: at a context threshold Claude updates the docs and writes a handoff, the session compacts with its instructions, then a kickoff prompt carries on. Gives Claude a live context reading and a /wrap-up skill

Shows how long the prompt cache stays warm and suggests compacting a large context at the moments where it pays off.

The fix for AI context overload isn't a bigger window, it's a trip wire. One Claude Code mod kills the session at 85% full and restarts clean, because past that point the model doesn't fail loudly. It just starts hallucinating quietly.