When this matters
- The same build or deployment warning appears in several memory sections.
- The agent spends tokens on old exploratory notes before reading the current task.
- A repo has grown multiple overlapping instruction files.
Claude Code context bloat
Claude Code Context Bloat Scanner
Claude Code context bloat happens when useful continuity turns into a long pile of repeated, outdated, or low-priority notes. The fix is not deleting everything. The fix is classifying which context still changes behavior and compressing it into a smaller brief.
How to run the audit
- Measure repeated phrases, dead commands, outdated dates, and low-signal sections.
- Identify the few memories that should remain pinned.
- Convert noisy history into a concise current-state brief.
- Move incident-only details into expiring notes.
- Verify that the final brief still covers safety and workflow rules.
Common risks
- Removing the wrong context can erase important safety constraints.
- Keeping everything can make the agent ignore the highest-priority instruction.
- Bloat often hides contradictions until a production change depends on them.
How Memory Hygiene Audit connects this to checkout
Memory Hygiene Audit scores bloat and produces a cleanup receipt that separates durable context from transcript residue.
Teams can preview the score, then use the Team Hygiene annual checkout to generate the full cleanup diff, reviewer notes, and agent-readable JSON policy.