memory-files-tools

claude-memory-files

Curated, modular memory and instruction files for Claude Code, maintained by Ribose for developers across our orgs (riboseinc/*, metanorma/*, lutaml/*, plurimath/*, relaton/*).

📖 Documentation site: riboseinc.github.io/claude-memory-files — getting started, concepts (scope rubric, mechanism/data, hooks-deferred rationale), task guides (cherry-pick via curl, slash commands, submitting your own file, customising personal-share, updating installed files), reference (frontmatter schema, file types, validator rules, manifest format, slash commands), and roadmap.

What this repo is

A flat, cherry-pickable collection of small files that you can drop into your local ~/.claude/ setup individually — not a monolithic configuration to adopt wholesale. Each file declares mandatory YAML frontmatter (name, description, type, scope, …) and lives under one of five category directories that mirror Claude Code's content surface:

• instructions/ — files that land at ~/.claude/instructions/<name>.md and get @-included from your ~/.claude/CLAUDE.md. Load every session.
• memory/ — files that land at ~/.claude/memory/<name>.md. Reference, not auto-loaded.
• settings-fragments/ — JSON snippets that deep-merge into ~/.claude/settings.json (especially permission allowlists).
• project-claude-md/ — per-project-type CLAUDE.md starter templates installed into a chosen project directory.
• rules/ — Anthropic's native .claude/rules/ files with paths: frontmatter; load only when matching files are in scope.

A sixth category, hooks/, is reserved for v2 — see SAFETY.md for the policy.

Two install paths are planned: a curl | bash one-liner per file (zero install ceremony), and a small companion plugin shipping slash commands (/install-memory-file, /list-memory-files, etc.). Both write to a user-side manifest at ~/.claude/.memory-files-manifest.json. See CONTRIBUTING.md for the contribution flow and SCHEMA.md for the frontmatter contract.

What this repo deliberately does NOT host

Slash commands, subagents, skills, MCP server configurations, output styles, and keybindings are out of scope here. Those categories are abundantly catalogued already by mature aggregators (see "Further reading" below). Hosting them here would duplicate without adding value. We focus on the white-space categories the existing ecosystem doesn't cover: memory, instructions, settings fragments, project CLAUDE.md templates, and path-scoped rules.

Further reading and exemplar listings

Curated outbound links to aggregator listings that stay maintained — not individual exemplar files that bit-rot the day after they're written:

• hesreallyhim/awesome-claude-code — the canonical community-maintained awesome- aggregator
• Anthropic's memory documentation — authoritative on CLAUDE.md semantics, @-imports, .claude/rules/, and the MEMORY.md cap
• Anthropic's plugin docs and marketplace docs — relevant context for the install-tools plugin we ship
• anthropics/claude-plugins-official — Anthropic's curated marketplace (the "blessed" reference for plugin shape)
• davila7/claude-code-templates and the live aggregator at aitmpl.com — third-party template/component catalogue with active maintenance
• VoltAgent/awesome-claude-code-subagents — well-organised subagent catalogue (per-file YAML frontmatter, named categories)
• PatrickJS/awesome-cursorrules — cross-ecosystem analogue; cited deliberately because the cursor-rules failure mode is what our schema and validator exist to prevent

Design history

The foundational design is captured in issue #1 (seven comments covering distribution model, scope, schema, hooks-deferred, install paths, seed set, roadmap). The update-regulation policy is in issue #2. Implementation lands as a series of small PRs; see issue #1's roadmap comment for the sequence.

License

MIT.