claude-memory

claude-memory

A Claude Code plugin that keeps a repo's Claude Code memory layer healthy and under your control. It ships two skills:

Skill	Question it answers
/claude-memory:audit	Is the instruction/memory layer (CLAUDE.md, a root AGENTS.md, CLAUDE.local.md, .claude/rules/, auto-memory) healthy against official-doc criteria?
/claude-memory:stateless	Is Claude's auto memory on, where does it live, and how do I turn it off or wipe it?

The two skills split by axis: audit checks the health of the instruction/memory layer; stateless controls the on/off state and contents of the Claude-written auto-memory store. The configuration FILES, automation SET, and permission GRANTS are audited by the sibling skills in the separate claude-config plugin (audit, audit-automation-gaps, audit-permission-grants).

What the skills do

audit

Audits the files you write that shape Claude's behavior against a codified checklist derived from official Claude Code documentation (line budgets, deletion test, content placement, consistency, currency, auto-memory index integrity). A deterministic spine yields identical findings on identical repo state: the line budget counts a file with its @ imports expanded, the way the loader does; a nested AGENTS.md that a CLAUDE.md on its own path displaces, with no CLAUDE.md importing it, is reported as a file that never loads; the MEMORY.md index and orphan always-loaded rules are script-checked; and each finding on a standards-synced file routes its fix upstream rather than proposing an edit the next sync overwrites. The context-cost line is a bytes-per-token estimate over the always-loaded set and says so. Judgment-tier checks apply fixed criteria with model reading. Reports persist to the plugin's data directory. They audit contributor-personal auto-memory, so they never land in the repo.

Scope covers both layers that load every session: the project's CLAUDE.md (or the AGENTS.md and .claude/AGENTS.md read in its place) / CLAUDE.local.md / .claude/rules/, and the user-global ${CLAUDE_CONFIG_DIR:-~/.claude}/CLAUDE.md and ${CLAUDE_CONFIG_DIR:-~/.claude}/rules/. Every discovered file is tagged with its scope, so project-scoped criteria skip personal files instead of reporting a repo-scoped finding against one.

/claude-memory:audit          # audit (default)
/claude-memory:audit fix      # apply findings with per-item approval
/claude-memory:audit update   # refresh criteria from current official docs
/claude-memory:audit report   # show the last audit without re-running

stateless

Inspects and disables Claude Code auto memory, the notes Claude writes for itself per repo at ~/.claude/projects/<project>/memory/ (relocatable via autoMemoryDirectory). Scope is auto-memory only: the instruction layer (CLAUDE.md, a natively read AGENTS.md, .claude/rules/) belongs to audit, and transcripts / history are out of scope (Claude Code auto-cleans those via cleanupPeriodDays).

/claude-memory:stateless           # status (default) — effective on/off state + where the store lives
/claude-memory:stateless disable   # autoMemoryEnabled:false + CLAUDE_CODE_DISABLE_AUTO_MEMORY (scope-confirmed)
/claude-memory:stateless purge     # DESTRUCTIVE — delete auto-memory files after a confirmation gate

disable sets both the env var and the setting. The env var is authoritative and overrides autoMemoryEnabled per the env-vars doc; the setting is the persistent fallback. purge reads autoMemoryDirectory at every settings scope before it enumerates what to delete, shows a manifest, and deletes only after explicit confirmation. Claude Desktop / claude.ai account memory is a separate server-side store. The skill gives direction to the app's Settings → Memory controls rather than deleting it locally.

Consumer conventions

The skill reads the consuming repo's own CLAUDE.md / .claude/rules/ for project-specific instruction-layer policy: a team-shared-first codification rule, an always-loaded context-budget policy, or a documented exemption (e.g. a deliberate CLAUDE.md line-budget overage for a repo that runs a large rules layer). Such findings are surfaced under a REPO check-ID so they stay distinct from the doc-derived checks. Nothing project-specific is baked into the plugin.

Install

/plugin marketplace add melodic-software/claude-code-plugins
/plugin install claude-memory@melodic-software

Configuration

No userConfig. State: audit reports persist under the plugin's ${CLAUDE_PLUGIN_DATA} directory. They are contributor-local because they cover per-contributor auto-memory, so they never land in the consuming repo. Side effects: stateless disable edits a settings.json you choose (setting autoMemoryEnabled and an env var, then flagging a dotfile-manager backfill if the file is tracked); stateless purge deletes auto-memory *.md files after a confirmation gate. Both act only on the scope you confirm. Network: the audit update action fetches official docs pages (read-only). Scripts require git and standard shell utilities.

License

MIT (SPDX-License-Identifier: MIT).