basicmachines-co/basic-memory-skills
10 skills5.3k installs250 starsGitHub
Install
npx skills add https://github.com/basicmachines-co/basic-memory-skillsSkills in this repo
1Memory Notesmemory-notes is a Claude Code skill for authoring well-structured Basic Memory markdown files that parse into a searchable knowledge graph. Every note follows three sections: YAML frontmatter with title and tags, body content with semantic observations, and relations expressed as wiki-links between notes. Developers reach for memory-notes when creating architecture decisions, API notes, or research write-ups that agents must retrieve later during implementation. The skill encodes best practices for linking related concepts so Basic Memory can index categories and cross-references without custom tooling.735installs2Memory Reflectmemory-reflect is a background reflection skill inspired by sleep-time compute for coding agents. It reviews recent conversations and daily notes, extracts durable insights, and writes them into long-term memory instead of letting context decay between sessions. Developers reach for memory-reflect when an agent must remember decisions, coding preferences, and lessons learned across days or weeks of work. The skill is designed to run on cron or heartbeat triggers one to two times daily, or on explicit user request when reflection is needed immediately. It improves memory quality over time without interrupting active development sessions.663installs3Memory Defragmemory-defrag is a skill from basicmachines-co/basic-memory-skills that treats agent memory like a filesystem needing periodic defragmentation. It splits oversized files, merges duplicate entries, deletes stale information, and restructures the memory hierarchy when daily notes pile up without consolidation. Developers run it weekly or biweekly via cron, on demand when cleanup is requested, or when MEMORY.md exceeds roughly 500 lines. The skill keeps context windows efficient for Claude and similar agents that read markdown knowledge stores, preventing redundant or outdated notes from crowding out current facts during retrieval.534installs4Memory Tasksmemory-tasks is a task persistence skill from basicmachines-co/basic-memory-skills that manages work-in-progress using Basic Memory's schema system. Tasks are notes with type: Task that live in the knowledge graph, validate against a schema, and remain queryable after context compaction. Developers use memory-tasks when starting multi-step work with three or more steps, after compaction or restart to search for active tasks, and during pre-compaction flush to update all active tasks with current state. The skill bridges agent session limits and long-running implementation work without losing progress metadata.525installs5Memory Schemamemory-schema is a Basic Memory skill that manages structured note types through the Picoschema system. It discovers unschemaed notes, infers schemas, creates and edits schema definitions, validates notes, and detects drift so types like Task, Person, or Meeting stay uniform across a knowledge graph. Developers reach for memory-schema when agent memory must be queryable and validatable instead of freeform markdown, especially as new note patterns emerge and existing records need conformance checks before recall, search, or lifecycle automation runs.497installs6Memory Metadata Searchmemory-metadata-search is an AI & Agent Building skill for Basic Memory that lets agents find notes by structured frontmatter instead of free-text search alone. Any custom YAML key beyond the five standard fields—title, type, tags, permalink, and schema—is indexed as entity_metadata and queryable through the search_notes tool. The skill documents seven filter operators including $in, $gt, $gte, $lt, $lte, and $between, plus dot notation for nested keys like schema.version. Developers combine metadata_filters with optional text queries, or use tags and status shortcuts for filter-only searches. memory-metadata-search is the right skill when an agent must recall notes by status, priority, confidence thresholds, or invented custom fields without scanning entire conversation logs.487installs7Memory Lifecyclememory-lifecycle is a Basic Memory skill that governs how entities move through status stages in a folder-based knowledge graph for agent workflows. The core principle is archive, never delete: completed work moves out of active views but remains in the graph for future retrieval. The skill handles marking items done, finished, submitted, missed, or cancelled, moving entities between status folders, updating frontmatter, and resolving edge cases in multi-session tasks. Developers and agent operators reach for memory-lifecycle when Basic Memory context grows stale or cluttered and needs structured lifecycle management without losing historical context.479installs8Memory Literary Analysismemory-literary-analysis is a Claude Code skill from basicmachines-co/basic-memory-skills that turns novels, plays, poems, or non-fiction into a Basic Memory knowledge graph. The pipeline runs six phases—setup, seed, chapter processing, cross-referencing, validation, and visualization—using six schema types (Character, Theme, Chapter, Location, Symbol, LiteraryDevice) and wiki-linked notes organized under chapters/, characters/, themes/, and analysis/ directories. Developers reach for memory-literary-analysis when building book-club companions, teaching resources, or research projects that must survive across sessions and scale to roughly 200+ notes and 1000+ relations. Processing runs in batches of about ten chapters per pass, with schema_validate and canvas queries at the end.476installs9Memory Ingestmemory-ingest is a Basic Memory ingestion skill from basicmachines-co/basic-memory-skills that turns unstructured external input into structured knowledge graph entities. Developers invoke it when meeting transcripts, conversation logs, pasted documents, or email threads must be parsed, cross-referenced against existing Basic Memory notes, and written as linked entities with observations, relations, and action items. The skill extracts entities, searches for existing matches, proposes new entities with approval, and captures follow-up tasks. Reach for memory-ingest when raw information arrives in bulk and agents need durable, queryable memory instead of one-off chat context.475installs10Memory Researchmemory-research in basicmachines-co/basic-memory-skills researches external subjects via web search, synthesizes results, and creates structured Basic Memory entities with user approval. Explicit triggers include "Research [subject]", "Look up [subject]", "What do you know about [subject]?", and "Evaluate [subject]". Implicit triggers also activate on bare names like "Terraform" or URLs that imply a research request. Findings persist as durable memory for synthesis, citation, and follow-up questions across days or weeks instead of disappearing when a chat ends. Developers reach for memory-research when evaluating companies, people, technologies, or topics that require accumulated context the agent can revisit later. The workflow emphasizes structured entity creation rather than one-off summaries, making it suitable for ongoing diligence or technology evaluation. Agents compatible with Basic Memory use this skill whenever research output should become a reusable knowledge artifact.471installs