thedotmack/claude-mem
16 skills61.4k installs1M starsGitHub
Install
npx skills add https://github.com/thedotmack/claude-memSkills in this repo
1Mem SearchThe mem-search skill queries claude-mem persistent cross-session memory when users ask whether a problem was solved before or how prior work was done. It enforces a three-layer workflow for roughly tenfold token savings: search for an index of observation ids, timeline for chronological context around interesting hits, then get_observations for full detail on only the filtered ids. Step one uses the search MCP tool with query, limit, project, optional type and obs_type filters, and date ranges returning compact tables with ids, timestamps, types, and titles. Step two calls timeline with anchor or query plus depth_before and depth_after to interleave observations, sessions, and prompts around a discovery. Step three batch-fetches selected observations in one request instead of many individual calls. Invoke for questions like did we already fix this, how did we solve X last time, or what happened last week. Optional knowledge agents synthesize conversational answers from observation history when raw records are not enough.7kinstalls2Smart ExploreThe smart-explore skill overrides default agent exploration with a three-layer tree-sitter workflow: smart_search ranks symbols across directories, smart_outline returns structural skeletons, and smart_unfold fetches complete symbol source on demand. While active, agents should prefer these MCP tools over Read, Grep, and Glob for code files above roughly one hundred lines. The core principle is index first, fetch on demand: map structure before loading implementation details. smart_search walks directories, parses code files, and returns ranked symbols with folded file views in one call, replacing the Glob to Grep to Read discovery cycle. smart_outline exposes functions, classes, methods, imports, and properties per file. smart_unfold returns full symbol bodies with guaranteed AST node boundaries. Documented token economics show four to eight times savings on file understanding and eleven to eighteen times versus Explore agents. Bundled tree-sitter grammars cover JavaScript, TypeScript, Python, Go, Rust, Ruby, Java, and C-family languages, with custom grammar registration via .claude-mem.json.5.8kinstalls3Make PlanThe make-plan skill turns an orchestrator agent into a planner that writes LLM-friendly phased implementation plans grounded in real documentation rather than assumed APIs. Phase 0 always deploys Documentation Discovery subagents to read docs, examples, and existing patterns, then consolidate an Allowed APIs list with anti-patterns before any implementation phases are authored. Each later phase frames copy-from-docs tasks with specific file and line references, verification checklists, and guards against invented methods or undocumented parameters. Subagents must follow a reporting contract listing sources consulted, concrete findings, copy-ready snippet locations, and confidence gaps; the orchestrator rejects conclusions without evidence. The final verification phase checks implementations against documentation, greps for known bad patterns, and runs tests. Plans are structured so each phase can run in a new session with self-contained references, pairing naturally with the sibling do skill for execution. Use when asked to plan a feature, multi-step task, or implementation before running do.5.8kinstalls4DoThe do skill makes an orchestrator agent execute phased plans by deploying subagents for all implementation work rather than coding directly. Each phase assigns one clear objective per subagent and requires evidence such as commands run, outputs, and files changed before advancing. Implementation subagents must copy patterns from documentation, cite unfamiliar API sources in code comments, and stop to verify when an API appears missing instead of inventing it. After each phase, separate Verification, Anti-pattern, and Code Quality subagents run the plan checklist, grep for known bad patterns, and review changes; a Commit subagent runs only after verification passes. Between phases, a Branch or Sync subagent pushes the working branch and prepares handoff context so the next phase can start fresh while retaining plan references. Fresh subagents are used when context is large or unclear. The skill pairs with make-plan for documentation-first planning and prevents skipped verification, undocumented parameters, and premature commits. Use when asked to execute, run, or carry out a plan created by make-plan.5.6kinstalls5Timeline ReportThe timeline-report skill generates comprehensive narrative analysis titled Journey Into Project from claude-mem persistent memory timelines. Prerequisites require a running claude-mem worker with port resolved from CLAUDE_MEM_WORKER_PORT, settings.json, or per-UID fallback 37700 plus uid mod 100. Worktree detection maps git worktrees to parent project names for correct API queries. Step one fetches the full timeline via curl to /api/context/inject?project=NAME&full=true, then estimates token count and confirms with users when exceeding 100K tokens. A subagent analyzes the entire timeline plus SQLite queries at ~/.claude-mem/claude-mem.db for Token Economics and Memory ROI sections. Required report sections cover genesis, architectural evolution, breakthroughs, work patterns, technical debt, debugging sagas, memory continuity, token economics, timeline statistics, and meta lessons. Output saves to ./journey-into-PROJECT_NAME.md by default targeting three thousand to six thousand words with observation IDs and timestamps cited throughout the narrative.5.3kinstalls6Knowledge AgentThe knowledge-agent skill Build and query AI-powered knowledge bases from claude-mem observations. Use when users want to create focused "brains" from their observation history, ask questions about past work patterns, or compile expertise on specific topics. Build and query AI-powered knowledge bases from claude-mem observations. Knowledge agents are filtered corpora of observations compiled into a conversational AI session. Build a corpus from your observation history, prime it (loads the knowledge into an AI session), then ask it questions conversationally. Think of them as custom "brains": "everything about hooks", "all decisions from the last month", "all bugfixes for the worker service". ``text build_corpus name="hooks-expertise" description="Everything about the hooks lifecycle" project="claude-mem" concepts="hooks" limit=500 `` text build_corpus name="hooks-expertise" description="Everything about the hooks lifecycle" project="claude-mem" concepts="hooks" limit=500 Filter options: - project — filter by project name - types — comma-separated: decision, bugfix, feature, refactor, discovery, change - concepts — comma-separated concept tags - files — comma-separated file paths.4.3kinstalls7Learn Codebaselearn-codebase is an agent skill that front-loads deep project familiarity by systematically reading every source file in full before coding starts. It triggers when users ask to learn the codebase, read the codebase, prime, or get up to speed on a new project. The workflow is explicit and non-negotiable: read each source file completely, using Read tool offset and limit parameters to page through large files in chunks such as 500-line windows. The skill frames this as building a cognitive cache that reduces costly rediscovery later, and asks reviewers to weigh token cost against long-project savings before warning users. Developers reach for it at the start of unfamiliar repositories, after large refactors, or whenever downstream tasks depend on accurate mental models of file layout, conventions, and cross-module dependencies. It does not substitute targeted debugging or incremental exploration when only one subsystem matters, but it excels when broad codebase literacy is the bottleneck before feature design or multi-file changes.3.8kinstalls8Pathfinderpathfinder is an orchestrator skill that maps a codebase into feature-grouped flowcharts, hunts duplicated concerns across features, proposes a unified architecture, and emits per-system /make-plan handoff prompts without writing implementation code. Phase 0 deploys one Feature Discovery subagent to propose boundaries from the source tree, README, and import graph, writing 00-features.md before any fan-out. Phase 1 runs one Flowchart subagent per feature in parallel, producing Mermaid flowchart TD diagrams with every node labeled by file:line plus side effects and external dependencies. Phase 2 runs within-feature and cross-feature duplication subagents, requiring every claim to cite at least two file:line locations in 02-duplication-report.md. Phase 3 has the orchestrator write 03-unified-proposal.md, rejecting unnecessary abstraction layers, dual-path feature flags, and registry factories when a switch suffices. Phase 4 outputs copy-pasteable /make-plan prompts in 04-handoff-prompts.md listing consolidated entry points and exact call sites to rewrite. Subagents must report sources consulted, concrete findings, diagrams, and confidence notes; the orchestrator redeploys reports la.3.7kinstalls9How It Workshow-it-works is an agent skill that explains the claude-mem persistent memory system for Claude coding sessions. Every Read, Edit, and Bash action becomes a compressed observation that is summarized at session end, and relevant observations auto-inject into future prompts so the next session starts with prior context without re-explaining the codebase or rediscovering decisions. Memory injection begins on the second session in a project because the first session seeds memory while subsequent sessions receive auto-injected context for relevant past work. Optional front-loading is available through the learn-codebase command to ingest an entire repository in one pass taking roughly five minutes. All data stays local under the home directory claude-mem folder including SQLite database, vector index, logs, and settings, removed cleanly on npx claude-mem uninstall. The only external calls are to whichever AI provider is configured for compression such as Claude, OpenRouter, or Gemini. Developers reach for it when users ask how claude-mem works, what the tool is doing, when injection kicks in, or where memory data is stored on disk.3.6kinstalls10BabysitBabysit PR keeps watching a pull request until it is actually clean, not after a single status check. Workflow identifies PR number, branch, and base, confirms non-draft state, and inspects mergeability, checks, review decision, comments, and review threads. Pending checks poll on a thirty to sixty second cadence unless the user requests otherwise. New comments and unresolved threads are read with bot summaries verified against code. Real issues get focused commits, relevant tests, push, and loop back to monitoring. Stale review threads resolve only after verifying the latest head addresses the feedback. GitHub CLI gh pr view supplies coarse status while GraphQL paginates unresolved review threads with jq filters. Stop only when checks pass or are intentionally skipped, review decision is acceptable, no actionable comments remain, and no unresolved threads remain.3.4kinstalls11WowerpointThe wowerpoint skill converts exactly one source document into a shareable NotebookLM slide-deck PDF using kawaii narrative styling. Videos and podcasts from the same engine are out of scope. Setup installs notebooklm-py with Playwright chromium and requires jq; users authenticate interactively via notebooklm login. Workflow ensures the source doc is comprehensive enough, runs auth pre-flight, creates a notebook and adds the single source, then spawns a background subagent for ten-minute generation. The subagent waits for source readiness, generates slide-deck with a one-sentence kawaii prompt, waits for artifact, downloads to adjacent stem-slides.pdf, and optionally uploads to WOWerpoint Server for a mobile share URL when env vars are set. Parent prints the NotebookLM live URL and ends the turn while generation runs. JSON envelope keys differ per CLI command; wrong keys cause silent empty-string failures. Weak sources should be expanded via mem-search before deck creation.3.1kinstalls12Design Isdesign-is is a Claude Code skill that audits an existing design against Dieter Rams' ten 'Good design is' principles and turns the critique into a plan handoff. It runs as an orchestrator: after locking scope in 00-scope.md, it fans out evidence subagents - structural, visual, copy and honesty, weight and friction, and optionally accessibility - that return cited facts such as interactive-element counts, spacing and type scales, contrast ratios, dark-pattern flags, and initial JS bytes. The orchestrator alone scores each principle 0-3 against fixed per-principle anchors, applying tie-breaker and score-worst rules, then reads the verdict mechanically: REFINE at a total of 20 or more with no zero score, REDESIGN below 20 or when a load-bearing principle scores 0, NEW when no real artifact exists yet. Developers use it when a UI needs a design review that should end in a ready-to-run /make-plan prompt, with five markdown artifacts documenting scope, evidence, scorecard, verdict, and handoff.2.8kinstalls13Weekly DigestsThe weekly-digests skill Generate a serial week-by-week narrative digest of a project's full claude-mem timeline. Splits the timeline into per-ISO-week files, then runs one consecutive subagent per week - each receiving the prior week's carry-forward block - to produce one chapter per ISO week of data. Use when asked for "weekly digests", "week-by-week story", "serial timeline", or "narrative chapters" of a project's history. It covers "Story chapters of project ". Key workflows include "Run a digest for each week". Trigger when the user asks for: - "Weekly digests" - "Week-by-week story" - "Serial timeline" - "Story chapters of project " - "Run a digest for each week" - "Continue the story week by week" If the user wants a single sweeping report, use timeline-report instead. This skill is for serial chapter format. Developers invoke weekly-digests when the task matches the triggers and reference files in SKILL.md for grounded, stepwise execution.2.7kinstalls14Oh My IssuesThe oh-my-issues skill Cluster a GitHub issue backlog by root cause into a small set of plan-master issues, redirect children with a standardized comment, and bundle architectural-fix PRs that close clusters atomically. Use when an issue tracker has accumulated dozens of reports that share underlying defects, when asked to triage / consolidate / cluster / dedupe issues, when asked to build a plan series or roadmap from open issues, or when routing a new incoming bug into an existing plan. It covers the repo has 20+ open issues and many feel like duplicates or platform-specific symptoms of the same defect.. Key workflows include the user asks to "triage", "consolidate", "cluster", "dedupe", "group", or "make a plan from" the issue list.. - The repo has 20+ open issues and many feel like duplicates or platform-specific symptoms of the same defect. - The user asks to "triage", "consolidate", "cluster", "dedupe", "group", or "make a plan from" the issue list. - A new bug is filed and the user wants to know whether2.7kinstalls15Standupstandup is an agent skill from thedotmack/claude-mem that coordinates multiple coding agents—each embodying a git branch or PR—in one shared markdown file acting as a group chat room. A facilitator runs rounds, schedules speakers, and decides when the standup is complete; participating agents read the room, state their case, react, and return on recall rather than looping indefinitely. The goal in front matter is typically collapsing everyone's work into one consolidated worktree. Agents interact through a standup.mjs CLI with read and speak commands supplied by the facilitator. Use standup when parallel agent branches produce conflicting plans, when several PRs need a negotiated merge strategy, or when autonomous agents must reach consensus before consolidating code. The skill treats standup as conversation—not a form—so agents push back, change positions, and land on one plan together.2kinstalls16Claude Memclaude-mem is a Claude Code skill in the AI & Agent Building category. Persistent memory system for Claude Code - context compression across sessions0installs