npm package discovery and stats viewer.

Discover Tips

  • General search

    [free text search, go nuts!]

  • Package details

    pkg:[package-name]

  • User packages

    @[username]

Sponsor

Optimize Toolset

I’ve always been into building performant and accessible sites, but lately I’ve been taking it extremely seriously. So much so that I’ve been building a tool to help me optimize and monitor the sites that I build to make sure that I’m making an attempt to offer the best experience to those who visit them. If you’re into performant, accessible and SEO friendly sites, you might like it too! You can check it out at Optimize Toolset.

About

Hi, 👋, I’m Ryan Hefner  and I built this site for me, and you! The goal of this site was to provide an easy way for me to check the stats on my npm packages, both for prioritizing issues and updates, and to give me a little kick in the pants to keep up on stuff.

As I was building it, I realized that I was actually using the tool to build the tool, and figured I might as well put this out there and hopefully others will find it to be a fast and useful way to search and browse npm packages as I have.

If you’re interested in other things I’m working on, follow me on Twitter or check out the open source projects I’ve been publishing on GitHub.

I am also working on a Twitter bot for this site to tweet the most popular, newest, random packages from npm. Please follow that account now and it will start sending out packages soon–ish.

Open Software & Tools

This site wouldn’t be possible without the immense generosity and tireless efforts from the people who make contributions to the world and share their work via open source initiatives. Thank you 🙏

© 2026 – Pkg Stats / Ryan Hefner

fractal-agentic

v2.6.6

Published

Progressive one-boss discovery, /orchestrate delivery runtime, and vendored skills for AI coding agents.

Readme

Fractal Agentic

The coding-agent plugin for serious delivery.
An orchestrator prime - charioteer: domain bosses, a single orchestrator, and a vendored armory of skills, agents, and commands.

Why install this

| Without Fractal Agentic | With Fractal Agentic | |---|---| | Ad-hoc process every session | /orchestrate delivery runtime | | Unclear who “owns” UI vs security vs scaffold | 7 executive bosses with missions and handoffs | | Skills scattered or missing | Vendored armory under skills/ (no runtime symlinks) | | Users must @ the plugin | AGENTS mandate — detect and use automatically | | Pin install freezes work | Non-blocking pins — product work always proceeds |

Stack defaults: Svelte 5 + SvelteKit + indented SASS (Tauri when desktop). Works in any project once the mandate is pasted.

| Pillar | Entry | |---|---| | Identity | SOUL.md | | Startup router + boss playbooks | AGENTS.mdBosses | | Orchestrator | /orchestrate · boss-orchestration | | Armory indexes | skills/INDEX.md · agents/INDEX.md · commands/INDEX.md | | Hooks (optional) | hooks/README.md | | Customize | CUSTOMIZE.md | | Pin policy | docs/progression.md | | Troubleshooting | TROUBLESHOOTING.md · docs/troubleshooting.md |

This directory is the plugin root — marketplace install unit, FRACTAL_AGENTIC_ROOT, and all manifests live here.
Host-agnostic content with thin adapters (CLAUDE.md, GEMINI.md, KIMI.md, OPENCODE.md, optional hooks).

What lives where (this package vs the rest of the repo)

| Kind | Location | Audience | | --- | --- | --- | | Agent identity and startup | SOUL.md, AGENTS.md, selected boss playbook, host shims | Agents must load | | Armory | skills/, agents/, commands/ | Agents + users | | Shipped support docs | docs/ | Dual — offline humans and agents (docs/INDEX.md labels each page) | | Website | sibling ../site/ | Humans only; renders this package, does not replace it | | Repo shell | parent root | Clone, credits, marketplace catalog — see ../LAYOUT.md | | Doc SSOT rules | docs/doc-ownership.md | Router vs nested boss playbooks; agent vs dual vs site |

Full documentation: docs/INDEX.md.
Browse online: fractal-agentic.vercel.app · Docs hub

Contents: Install · CLI health & sync · Auto-use · Quick start · Bosses · Orchestration · Layout · Customize · Docs KB


Install

Full step-by-step for every host: docs/02-install.md
(manifest map, failures, update commands).

Universal NPX Installer (Recommended)

# Auto-detect all host environments (Antigravity, Claude Code, Codex) and install
npx fractal-agentic install

# Or target a specific host
npx fractal-agentic install --target=antigravity

Claude Code (Marketplace / Git)

# Add the repository as a plugin marketplace
claude plugin marketplace add fractalmandala/fractal-agentic

# Install the plugin
claude plugin install fractal-agentic@fractal-agentic

Or from local checkout:

cd /path/to/fractal-agentic
claude plugin marketplace add .
claude plugin install fractal-agentic@fractal-agentic
export FRACTAL_AGENTIC_ROOT="$PWD/plugin"

One session only: claude --plugin-dir /path/to/fractal-agentic/plugin

Then: new Claude session. Confirm with claude plugin list (fractal-agentic@fractal-agentic enabled).

Codex

codex plugin marketplace add fractalmandala/fractal-agentic \
  --sparse .agents/plugins \
  --sparse plugin

Enable fractal-agentic in the Codex plugin UI · new task.
Catalog file: repo .agents/plugins/marketplace.json (not the Claude marketplace file).

Google Antigravity

npx fractal-agentic install --target=antigravity

Or copy/symlink ./plugin to ~/.gemini/config/plugins/fractal-agentic.

Git / any host

git clone --filter=blob:none --sparse https://github.com/fractalmandala/fractal-agentic.git
cd fractal-agentic && git sparse-checkout set plugin .agents .claude-plugin
export FRACTAL_AGENTIC_ROOT="$PWD/plugin"
sh "$FRACTAL_AGENTIC_ROOT/scripts/resolve-plugin-root.sh"

Then register a marketplace (Claude/Codex above) or paste the AGENTS snippet and point the host at $FRACTAL_AGENTIC_ROOT.

Cursor / Gemini / Kimi / OpenCode

See docs/02-install.md — AGENTS snippet + FRACTAL_AGENTIC_ROOT + host shim (GEMINI.md / KIMI.md / OPENCODE.md).

After install (all hosts)

  1. Paste project-integration/AGENTS-SNIPPET.md into each project’s AGENTS.md (or run npx fractal-agentic install --project).
  2. Optional: /hooks-init · /improve-init · /wiki-init
  3. Optional pins (TOML hosts): sh scripts/install-agents.sh
  4. Health: sh scripts/check-armory.sh

Missing pins → keep shipping with pins: unverified (progression.md).


CLI health & sync

The installer CLI (npx fractal-agentic or fractal-agentic when linked) also keeps installed copies healthy:

# Armory health + host-copy drift check (exit 1 when unhealthy)
fractal-agentic doctor              # fa-doctor · --json for machines

# Sync every host copy to the root version
fractal-agentic update              # fa-update · --dry-run · --force · --target=<host|all>

doctor statuses: ok · drift · missing · symlink · fail — healthy means the armory check passes and every detected host copy is ok, symlink, or absent. update refreshes drifted copies (Antigravity, Claude Code, Codex, Agents) and never overwrites symlinks or matching versions. Full flag reference: bin/README.md.


Auto-use from any project (AGENTS.md mandate)

So models do not need @fractal-agentic, paste the mandate into each project’s AGENTS.md.

| File | Purpose | |---|---| | project-integration/AGENTS-SNIPPET.md | Copy-paste block | | scripts/resolve-plugin-root.sh | Probe: prints a complete router + boss-playbook + runtime root or exits 1 |

Per-project recipe

  1. Paste project-integration/AGENTS-SNIPPET.md near the top of the project’s AGENTS.md.
  2. Set env to this plugin root:
    export FRACTAL_AGENTIC_ROOT=/Users/you/fractal-agentic/plugin
  3. Confirm:
    sh "$FRACTAL_AGENTIC_ROOT/scripts/resolve-plugin-root.sh"
  4. Agents: detect → read the AGENTS.md router → select and read exactly one nested boss INDEX.md → stop discovery → load boss-orchestration only for delivery work. If not found: say so once; continue with project rules. Never block product work.

Conflict rule: project-local conventions win for this repo’s code; the plugin owns process, armory, and delivery guidance.


Quick start

1. Optional capability agents (one-time, host-dependent)

Plugin install does not auto-register custom-agent TOML files. On hosts that use them:

# from this plugin root (…/fractal-agentic/plugin)
sh scripts/install-agents.sh
# or explicit target (example)
sh scripts/install-agents.sh --target-dir "$HOME/.codex/agents"

Then, if you want pins discovered in the spawn catalog, start a new agent session/task (optional — work must not wait):

| Lane | agent_type | Typical pin (host-specific) | |---|---|---| | Routine implementer | fractal_agentic_routine_implementer | high-throughput implementer | | Complex implementer | fractal_agentic_complex_implementer | high-reasoning implementer | | Fresh reviewer | fractal_agentic_fresh_reviewer | independent review / read-only when available |

sh scripts/install-agents.sh --check

2. Health checks

sh scripts/check-armory.sh   # orchestration assets + critical skills
sh scripts/verify.sh         # full package: TOML pins, installer, runtime inspector

3. Use in a session

| You want… | Do this | |---|---| | Ship a feature end-to-end | /orchestrate (or load boss-orchestration skill) | | Work in a domain | /activate-boss-svelte (or design/code/agent/creator/workflow/meta) then /orchestrate | | Browse the one-boss path | Read AGENTS.md then Domain bosses | | Live inventories | skills/INDEX.md · commands/INDEX.md · agents/INDEX.md |

Typical flow

1. User request
2. /orchestrate  →  select ACTIVE BOSS (domain)
3. Preflight: install-agents.sh --check + spawn types present
4. Write five-part spec (objective, ownership, interfaces, constraints, verification)
5. Spawn routine OR complex implementer
6. Primary session re-verifies diff + commands (worker reports are claims)
7. Spawn fresh reviewer → ship | fix-first | rethink
8. ship only → report done; fix-first → re-delegate + new review
9. Release-critical → /santa-loop after ship

Mental model

┌─────────────────────────────────────────────────────────────┐
│  PRIMARY SESSION (architect)                                 │
│  Intent · boss selection · specs · verify · accept           │
└───────────────────────────┬─────────────────────────────────┘
                            │
         ┌──────────────────┼──────────────────┐
         ▼                  ▼                  ▼
   Domain axis         Capability axis     Ship gate
   (router + boss)     (TOML pins)         (reviewer)
   Design/Code/…       routine/complex     ship|fix-first|rethink
  • AGENTS.md = startup router (what to read first, one-boss selection, and handoffs).
  • Nested boss INDEX.md = mission, mapped skills/agents/commands, phases, and verification defaults.
  • boss-orchestration = runtime (what happens when, which model lane, when “done” is legal).
  • activate-boss-* = load the router plus one domain’s authoritative playbook.
  • Domain agents (svelte-reviewer, security-reviewer, …) = specialists; optional consults. They do not replace the fresh-reviewer verdict under orchestration.

Domain bosses

| Boss | Activate | Owns | |---|---|---| | Design | /activate-boss-design | Tokens, UI craft, a11y, motion, visual QA (Svelte-first) | | Code | /activate-boss-code | Security, tests, debt, perf, docs-from-code, santa-loop release | | Agent | /activate-boss-agent | Product agent harness/memory/eval/MCP (not personal OS) | | Svelte | /activate-boss-svelte | Runes, Kit data-flow, SASS, port-component lane | | Creator | /activate-boss-creator | Scaffold → ship; executive armory pull | | Workflow | /activate-boss-workflow | Personal habits, prune, context-save/restore | | Meta | /activate-boss-meta | ECC install, skill stocktake/comply/promote/prune |

Decision tree (short):

UI craft / a11y / motion?           → Design
Svelte / Kit / port?                → Svelte
Security / tests / debt / docs?     → Code
Product agents / harness / MCP?     → Agent
New app/site/package?               → Creator
Personal loops / cost / instincts?  → Workflow
ECC portfolio / skill health?       → Meta
Unclear?                            → Creator (or Workflow to observe)

Full handoffs and boss boundaries: Domain bosses.


Orchestration runtime (detail)

Skill package

skills/boss-orchestration/
  SKILL.md
  agents/openai.yaml          # Codex skill UI (display + default prompt)
  references/
    role-contracts.md         # spawn shapes, five-part specs, review packets
    routing-matrix.md         # domain × capability × stack
    handoffs.md
    boss-prompts.md           # per-boss CONSTRAINTS injects

Five-part implementation contract

Every worker prompt must include:

  1. OBJECTIVE — observable outcome
  2. FILES AND OWNERSHIP — exact owned paths; concurrent-edit rules
  3. INTERFACES — compatibility that must not break
  4. CONSTRAINTS — repo rules + ACTIVE BOSS bullets from boss-prompts.md
  5. VERIFICATION — exact commands + expected evidence

Workers return an IMPLEMENTATION REPORT. Primary re-runs verification; the report alone is not evidence.

Routing

| Task shape | Lane | |---|---| | Spec largely determines the result | fractal_agentic_routine_implementer | | Judgment, security, concurrency, wide blast | fractal_agentic_complex_implementer | | Misclassified after one routine attempt | Correct spec → escalate complex | | Pre-architecture decision | Fresh reviewer consult | | After primary verification | Fresh reviewer final → ship | fix-first | rethink |

Preflight rules (Codex)

  1. install-agents.sh --check exits 0 (byte-exact templates).
  2. Spawn tool exposes all three exact agent_type names.
  3. Observed model/effort match pins (public details first; else inspect-agent-runtime.sh <thread-id>).
  4. Reviewer: report observed sandbox + permission profile; do not claim read-only unless observed.

No silent fallback to another role, model, or effort.

Non-Codex hosts

Keep contracts, routing, and ship gate. State that model pins are unverified. Use domain agents for specialist review. Completion still requires an explicit ship|fix-first|rethink packet.


Scripts

| Script | Purpose | |---|---| | scripts/install-agents.sh | Install or --check the three TOML custom agents | | scripts/inspect-agent-runtime.sh | Allowlisted routing JSON from a subagent thread id | | scripts/check-armory.sh | Core files + critical skill path health | | scripts/verify.sh | Full local verification suite |

Installer behavior matches Sol Advisor: never overwrites a differing destination file; never mutates config.toml.


Layout

This directory is the plugin (marketplace install unit / FRACTAL_AGENTIC_ROOT).

plugin/                     ← INSTALL / FRACTAL_AGENTIC_ROOT (agents load this tree)
  SOUL.md · AGENTS.md       # identity + startup router (required)
  CLAUDE.md · GEMINI.md …   # host shims
  README.md · TROUBLESHOOTING.md
  plugin.json · .claude-plugin/ · .codex-plugin/
  skills/ · agents/ · commands/
  hooks/ · workflows/ · scripts/
  project-integration/      # AGENTS mandate snippet
  docs/                     # dual support docs (ship with package; site also renders them)

Repo siblings: ../site/ (website), ../LAYOUT.md (root + plugin + site rules).


Vendored skills (no external symlinks)

All skills live as real directories under skills/. Fractal Agentic does not depend on agentic/comparisionset, agentic/skills, curated-curor, or component-porter at runtime. Those trees were copy sources only.

| Source used at vendor time | Role | |---|---| | Outer packs, curated-curor, and port-component | Copy sources for selected skills | | High-value ECC/comparisionset subset | Copy source for delivery, product, and content skills | | Native plugin ECC skills | Core armory source | | Live inventory | See skills/INDEX.md; all entries are vendored locally |

High-value ECC subset inside this plugin

| Cluster | Skills | Natural boss | |---|---|---| | Orch family | orch-pipeline, orch-build-mvp, orch-add-feature, orch-change-feature, orch-fix-defect, orch-refine-code, plan-orchestrate | Creator + /orchestrate | | Decide before build | council, intent-driven-development, product-lens, product-capability | Creator Phase 0 | | Agent depth | agent-architecture-audit, agent-introspection-debugging, ai-first-engineering, gan-style-harness, enterprise-agent-ops, iterative-retrieval, team-builder, parallel-execution-optimizer, mcp-server-patterns, dynamic-workflow-mode | Agent | | Code / ship | api-design, backend-patterns, hexagonal-architecture, git-workflow, github-ops, deployment-patterns, docker-patterns, canary-watch, code-tour, santa-method | Code | | Sites / content | content-engine, article-writing, seo, ui-demo, deep-research, documentation-lookup, knowledge-ops | Design / Creator / Workflow | | Misc | prompt-optimizer, search-first | Meta / Workflow |

How these relate to boss-orchestration

  • boss-orchestration = capability + domain runtime
  • orch-* = operation playbooks that can use the same lanes
  • plan-orchestrate = plan → orchestrate prompts
  • santa-method = skill form of dual review (/santa-loop is the command)

Adding more skills later

Copy (never symlink) into skills/<id>/, then map it in the owning nested boss playbook — see CUSTOMIZE.md.


Souls - What a soul is

Clawsouls / OpenClaw-style persona packs, not domain maps and not orchestration runtimes:

| File | Role | |---|---| | SOUL.md | Beliefs, how you work, boundaries, personality | | IDENTITY.md | Name, vibe, emoji | | STYLE.md | Voice, rhythm, anti-patterns (especially creative souls) | | HEARTBEAT.md | Periodic self-nudges (“check React/CSS updates”) | | AGENTS.md | One-line default agent prompt | | soul.json | Metadata, tags, recommended skills, tool allowlists |

Example souls in the set: frontend-dev (React-centric), graphic-designer, storyteller, scifi-writer.

How that differs from Fractal Agentic

| Layer | Fractal Agentic | Souls | |---|---|---| | Domain / armory | Bosses + nested boss playbooks | Soft “recommended skills” only | | Runtime / model pins | boss-orchestration + TOML | None | | Ship gate | Fresh reviewer + santa | None | | Voice / persona | Thin (boss-prompts = constraints) | Thick (beliefs, tone, style) | | Host | ECC / Codex / multi | OpenClaw / clawsouls ecosystem |

Shared armory (always available)

  • /orchestrate + boss-orchestration
  • /quality-gate
  • /security-scan + security-reviewer
  • /code-review + code-reviewer
  • /santa-loop (and vendored skill santa-method) for release-critical dual review