According to AlphaMoat, as of 2026-09, ForesigxtMemory has 2,685 downloads and 1 stars — ranked #3,959 of 5,078 AI skills overall, and #792 of 969 in AI Agent.
| Platform | Installs | Downloads | Stars | Comments |
|---|---|---|---|---|
| ClawHub ↗ | 83 | 2.6K | 1 | - |
| SkillHub ↗ | 146 | 2.7K | 1 | — |
Platforms count differently and update at different times.
Vaults may contain sensitive personal, project, or relationship data. Confirm the data scope and whether search sends content to external services. Migration and auto-linking can modify Markdown files, so use a dry run and backup first and retain a recoverable copy. Both documented search commands require qmd; review the CLI installation and vault location first.
Profile published 2026-10-29
Generated from the skill’s documentation as examples; not results of actual runs.
The counting method changed in 2026-09: skills with the same name are now counted separately by author, and the full catalog is covered. Dashed lines show months before the change; they are not compared with later months, and no growth is computed across the change.
| Month | Downloads | MoM | Stars |
|---|---|---|---|
| 2026-09 | 2.7K | — | 1 |
| 2026-08 | 2.5K | +4.7% | 1 |
| 2026-07 | 2.4K | +9.2% | 1 |
| 2026-06 | 2.2K | +5.4% | 1 |
| 2026-05 | 2.1K | — | 1 |
| 2026-03 | 1.7K | — | 1 |
The author’s original file, shown as plain text without edits or review by AlphaMoat. Links are not clickable; check commands before running them. Fetched 2026-10-23.
--- name: fsxmemory version: 1.3.1 description: Structured memory system for AI agents. Context death resilience (checkpoint/recover), structured storage, Obsidian-compatible markdown, and local semantic search. author: Foresigxt repository: https://github.com/Foresigxt/foresigxt-cli-memory --- # Foresigxt Memory
Structured memory system for AI agents. ## Install ```bash npm install -g @foresigxt/foresigxt-cli-memory ``` ## Setup ### Option 1: Initialize New Vault ```bash # Initialize vault (creates folder structure + templates) fsxmemory init ~/memory ``` ### Option 2: Use Existing Vault **For isolated workspace memory** (each workspace has its own vault): ```bash # Create .env in workspace root echo 'FSXMEMORY_PATH=/path/to/workspace/memory' > .env # All agents in THIS workspace use this isolated vault fsxmemory stats # Works automatically! ``` **For shared memory across all workspaces**: ```bash # Set global environment variable (in ~/.bashrc or ~/.zshrc) export FSXMEMORY_PATH=/path/to/shared/memory # All agents in ALL workspaces share the same vault ``` **Or**: Use `--vault` flag for one-time override: ```bash fsxmemory stats --vault /path/to/other/vault ``` ## Core Commands ### Store memories by type ```bash # Types: fact, feeling, decision, lesson, commitment, preference, relationship, project, procedural, semantic, episodic fsxmemory remember decision "Use Postgres over SQLite" --content "Need concurrent writes for multi-agent setup" fsxmemory remember lesson "Context death is survivable" --content "Checkpoint before heavy work" fsxmemory remember relationship "Justin Dukes" --content "Client contact at Hale Pet Door" fsxmemory remember procedural "Deploy to Production" --content "1. Run tests 2. Build 3. Deploy" fsxmemory remember semantic "Event Loop Concept" --content "JavaScript's concurrency model..." fsxmemory remember episodic "First Production Deploy" --content "Deployed v2.0 today, team was nervous but it went well" ``` ### Quick capture to inbox ```bash fsxmemory capture "TODO: Review PR tomorrow" ``` ### Search (requires qmd installed) ```bash # Keyword search (fast) fsxmemory search "client contacts" # Semantic search (slower, more accurate) fsxmemory vsearch "what did we decide about the database" ``` ## Context Death Resilience ### Checkpoint (save state frequently) ```bash fsxmemory checkpoint --working-on "PR review" --focus "type guards" --blocked "waiting for CI" ``` ### Recover (check on wake) ```bash fsxmemory recover --clear # Shows: death time, last checkpoint, recent handoff ``` ### Handoff (before session end) ```bash fsxmemory handoff \ --working-on "Foresigxt Memory improvements" \ --blocked "npm token" \ --next "publish to npm, create skill" \ --feeling "productive" ``` ### Recap (bootstrap new session) ```bash fsxmemory recap # Shows: recent handoffs, active projects, pending commitments, lessons ``` ## Migration from Other Formats Migrate existing vaults from OpenClaw, Obsidian, or other markdown-based systems: ### Analyze First (Dry Run) ```bash # See what would be changed without modifying files fsxmemory migrate --from openclaw --vault /path/to/vault --dry-run ``` ### Migrate with Backup ```bash # Recommended: Creates automatic backup before migration fsxmemory migrate --from openclaw --vault /path/to/vault --backup # The migration: # ✅ Adds YAML frontmatter to all markdown files # ✅ Renames directories (procedural→procedures, semantic→knowledge, episodic→episodes) # ✅ Creates .fsxmemory.json config file # ✅ Preserves all content and custom categories # ✅ Creates timestamped backup for rollback ``` ### Rollback if Needed ```bash # Restore from backup if something went wrong fsxmemory migrate --rollback --vault /path/to/vault ``` ### Migration Options ```bash # Available source formats --from openclaw # OpenClaw vault format --from obsidian # Obsidian vault format --from generic # Generic markdown vault # Migration flags --dry-run # Preview changes without modifying files --backup # Create backup before migration (recommended) --force # Skip confirmation prompts --verbose # Show detailed progress --rollback # Restore from last backup ``` ### Example: Migrate OpenClaw Vault ```bash # 1. Analyze first fsxmemory migrate --from openclaw --vault ~/.openclaw/workspace/memory --dry-run # 2. Run migration with backup fsxmemory migrate --from openclaw --vault ~/.openclaw/workspace/memory --backup --verbose # 3. Verify migration worked fsxmemory stats --vault ~/.openclaw/workspace/memory fsxmemory doctor --vault ~/.openclaw/workspace/memory ``` **Migration Speed**: ~53 files in 0.07 seconds ⚡ ## Auto-linking Wiki-link entity mentions in markdown files: ```bash # Link all files fsxmemory link --all # Link single file fsxmemory link memory/2024-01-15.md ``` ## Templates Reference Foresigxt Memory includes structured templates for consistent documentation. Location: `templates/` directory. ### Available Templates | Template | Type | Use For | Sections | |----------|------|---------|----------| | `decision.md` | decision | Key choices, architecture decisions | Context, Options, Decision, Outcome | | `procedure.md` | procedural | How-to guides, workflows, SOPs | Purpose, Prerequisites, Steps, Pitfalls, Verification | | `knowledge.md` | semantic | Concepts, definitions, mental models | Definition, Key Concepts, Examples, Why It Matters | | `episode.md` | episodic | Events, experiences, meetings | What Happened, Context, Key Moments, Reflection | | `person.md` | person | Contacts, relationships | Contact, Role, Working With, Interactions | | `project.md` | project | Active work, initiatives | Goal, Status, Next Actions, Blockers | | `lesson.md` | lesson | Insights, patterns learned | Situation, Lesson, Application | | `handoff.md` | handoff | Session continuity | Working On, Context, Next Steps, Blockers | | `daily.md` | daily | Daily notes, journal | Focus, Done, Notes | ### Template Usage Templates are automatically selected by memory type: ```bash fsxmemory remember decision "Title" --content "..." # → templates/decision.md fsxmemory remember procedural "Title" --content "..." # → templates/procedure.md fsxmemory remember semantic "Title" --content "..." # → templates/knowledge.md fsxmemory remember episodic "Title" --content "..." # → templates/episode.md fsxmemory remember relationship "Name" --content "..." # → templates/person.md fsxmemory remember lesson "Title" --content "..." # → templates/lesson.md ``` **To view template structure**: Read the template file in `templates/` directory before creating a memory document. **Template features**: - YAML frontmatter with metadata (title, date, type, status) - Structured sections with placeholder guidance - Wiki-link suggestions for connections - Auto-generated tags ## Folder Structure ``` vault/ ├── .fsxmemory/ # Internal state │ ├── last-checkpoint.json │ └── dirty-death.flag ├── decisions/ # Key choices with reasoning ├── lessons/ # Insights and patterns ├── people/ # One file per person ├── projects/ # Active work tracking ├── procedures/ # How-to guides and workflows ├── knowledge/ # Concepts and definitions ├── episodes/ # Personal experiences ├── handoffs/ # Session continuity ├── inbox/ # Quick captures └── templates/ # Document templates (9 types) ``` ## Best Practices 1. **Checkpoint every 10-15 min** during heavy work 2. **Handoff before session end** — future you will thank you 3. **Recover on wake** — check if last session died 4. **Use types** — knowing WHAT you're storing helps WHERE to put it 5. **Wiki-link liberally** — `[[person-name]]` builds your knowledge graph ## Integration with qmd Foresigxt Memory uses [qmd](https://github.com/tobi/qmd) for search: ```bash # Install qmd bun install -g github:tobi/qmd # Add vault as collection qmd collection add /path/to/vault --name my-memory --mask "**/*.md" # Update index qmd update && qmd embed ``` ## Configuration Foresigxt Memory supports three ways to set the vault path (in order of precedence): ### 1. Command-line flag (highest priority) ```bash fsxmemory stats --vault /path/to/vault ``` ### 2. Environment variable ```bash export FSXMEMORY_PATH=/path/to/memory fsxmemory stats ``` ### 3. .env file (for workspace-isolated memory) ```bash # Create .env in workspace root cat > .env << 'EOF' FSXMEMORY_PATH=/home/user/.openclaw/workspace/memory EOF # All fsxmemory commands in this workspace use this isolated vault fsxmemory stats fsxmemory checkpoint --working-on "task" ``` **Use .env when:** - ✅ **Isolating workspace memory** — Each project has its own separate vault - ✅ **Per-project configuration** — Different agents in different workspaces use different vaults - ✅ **Portable** — Workspace agents automatically use the right vault - ✅ **Git-safe** — Add `.env` to `.gitignore` to protect paths **Use global export when:** - ✅ **Sharing memory across workspaces** — All agents everywhere use one vault - ✅ **Centralized knowledge** — One source of truth for all projects **Environment Variables:** - `FSXMEMORY_PATH` — Vault path (can be set in shell or `.env` file) ## Publishing Skill Package To create a distributable skill package (includes SKILL.md and templates/): ```bash # Package the skill npm run package-skill # Output: dist-skill/fsxmemory-skill.zip (~8KB) ``` **Package contents:** - `SKILL.md` - Complete documentation and reference - `templates/` - All 9 memory templates - `.env.example` - Configuration template - `INSTALL.md` - Quick setup guide **Distribution:** Share the `fsxmemory-skill.zip` file with other agents/teams. They can extract it to get: - Complete skill documentation - Ready-to-use templates - Configuration examples **For OpenClaw/ClaudeHub:** The packaged skill is ready for upload to skill repositories. ## Links - npm: https://www.npmjs.com/package/@foresigxt/foresigxt-cli-memory - GitHub: https://github.com/Foresigxt/foresigxt-cli-memory - Issues: https://github.com/Foresigxt/foresigxt-cli-memory/issues
Records learnings, errors and corrections as Markdown files for later reuse and promotion into workspace memory.
Finds skills in local and remote catalogs from a plain-language request and recommends what to install.
A browser-automation command line skill for AI agents.
Turns corrections and self-review into tiered memory and promotes or archives it by use.
Reviews a third-party skill's source, permissions, and danger patterns against a checklist before it is installed.
Stores agent memory and cross-skill state as a typed, constraint-checked knowledge graph.
A workspace setup guide for planning proactive follow-up, working memory and review routines for an AI agent.
A guide to creating and updating effective Claude skills.
Data month: 2026-09 · Downloads, stars and installs are aggregated monthly from public skill registries (ClawHub, SkillHub). See methodology.