VerdictSwarm
v3.0.0Adds a fail-closed token-risk workflow using the current VerdictSwarm service contract.
Open source · Agent skills
Agent instructions should be inspectable, versioned, and testable.
Sentien Labs publishes versioned OpenClaw skill bundles with public source, release checks, and security-reviewed ClawHub distribution. Operators can inspect the exact instructions before installing them.
Catalog reviewed August 3, 2026.
Current catalog
Four focused skill bundles
Token Profiler
v1.1.0Retrieves a bounded, best-effort token profile for supported Solana and Base assets.
Agent Swarm Kit
v1.1.0Guides bounded multi-agent coordination with budgets, permissions, and stop rules.
Fleet Memory Manager
v1.1.0Maintains private-session memory with explicit consent, review-before-write, and deletion controls.
Public source
The GitHub repository is the reviewable source for files, tests, and release history.
Versioned registry
ClawHub distributes named releases while preserving the existing publisher and install path.
Parity checks
Release validation checks that clean installs match the intended source and reject retired service contracts.
Security and privacy
The instructions are part of the attack surface
Skills can influence which services an agent calls, what it stores, and when it acts. Releases therefore need the same review discipline as code: explicit permissions, bounded behavior, failure rules, changelogs, and reproducible source.
Memory-oriented tooling uses private-session-only defaults, explicit opt-in, review-before-write behavior, and retention/deletion controls. Shared channels must not silently load a personal profile.
A passing marketplace scan is useful evidence, not a permanent guarantee. Inspect the current source and requested permissions before installation.
Inspect source or install from ClawHub
GitHub is the public review and contribution surface. ClawHub is the versioned distribution surface for OpenClaw operators.