Skip to content

New Skills Discovered — June 26, 2026 (Afternoon Sweep)

Sources: skills.shfworks-tech/agenthood, letta-ai/letta-code Total new catalogued: 38 skills | Combined repos: 2 Date: June 26, 2026

Two major additions fill gaps in our agent-infrastructure and multi-agent categories: Agenthood provides a complete AI engineering team as skills.sh-compatible Markdown files, and Letta Code's full 24-skill harness (previously only 1 skill documented) brings stateful agent memory, MCP conversion, and self-improving infrastructure.


Skills Discovered

1. Agenthood — 14-Agent AI Engineering Team (fworks-tech/agenthood)

Agenthood packages 14 specialized AI agent roles as individual skills.sh-compatible SKILL.md files. Each agent is a self-contained role with responsibilities, standards, and communication patterns. Drop them into any agent runtime — Claude Code, Copilot, Hermes Agent, or the standalone TypeScript CLI.

GitHub: fworks-tech/agenthood · npm: agenthood (v3.0.0) · License: MIT

Agent Skill Role
🏗️ the-architect System design, ADRs, technical decisions
🔒 the-auditor Security scanning, vulnerability assessment, dependency audit
🐛 the-debugger Error triage, root cause analysis
🚪 the-doorman Validation, branch protection, CI health checks
🌐 the-envoy Cross-provider translation, convention validation
📦 the-herald Releases, versioning, changelog generation
📝 the-librarian Documentation, API references, knowledge management
🔮 the-oracle Institutional knowledge, authoring templates
🔍 the-reviewer Code review, standards enforcement
✍️ the-scribe Commits, PR descriptions, changelogs
👁️ the-sentinel Integrity monitoring, cross-member contradiction detection
🧭 the-steward Context economy, provider cache strategies
🧪 the-tester TDD, coverage analysis, edge case generation
⚖️ the-warden Code health, complexity enforcement

Why this matters for Hermes: Agenthood is a blueprint for multi-agent architecture. These 14 roles define clear boundaries, communication patterns, and escalation paths. Hermes agents can adopt individual roles (e.g., the-reviewer for PR workflows) or compose them into agent teams. The skills.sh compatibility means direct installation via npx skills add.


2. Letta Code — 24-Skill Stateful Agent Harness (letta-ai/letta-code)

Letta Code is a stateful agent harness built by the creators of MemGPT. Agents have persistent memory, identity, and learn over time through rewriting their own memory, skills, and prompts. Previously only 1 skill was catalogued — this update documents the full 24-skill suite.

GitHub: letta-ai/letta-code · npm: @letta-ai/letta-code · Built on: MemGPT research

Skill Category Description
initializing-memory Memory Set up agent memory blocks, identity, and personality
syncing-memory-filesystem Memory Git-backed memory sync (MemFS) across machines
finding-agents Discovery Search and discover existing agents in a workspace
migrating-memory Memory Migrate memory between agent runtimes and formats
creating-skills Skills Author new skills for agents programmatically
acquiring-skills Skills Discover and install skills from Hermes, ClawHub, GitHub
messaging-agents Communication Inter-agent messaging and channel management
adding-models Configuration Add and configure new LLM model backends
working-in-parallel Orchestration Dispatch and coordinate parallel agent workstreams
searching-messages Search Full-text search across all agent conversations
defragmenting-memory Memory Optimize and compact agent memory over time
migrating from codex and claude code Migration Migrate agents and workflows from other runtimes
converting-mcps-to-skills MCP Convert MCP servers into portable agent skills
dispatching-coding-agents Orchestration Spawn and manage coding sub-agents
scheduling-tasks Scheduling Cron jobs, heartbeats, and scheduled agent work
customizing-commands Configuration Add custom slash commands to the agent harness
Context Doctor Diagnostics Audit and repair agent context/memory quality
customizing-statusline UI Customize the agent terminal status display
creating-extensions Extensions Build harness extensions and plugins
modifying-the-harness Core Modify the agent harness itself for custom behavior
Plus 4 additional skills for channels, hooks, permissions, and secrets management

Why this matters for Hermes: Letta Code is the closest architectural cousin to Hermes Agent — both are stateful agent harnesses with memory, skills, and multi-agent orchestration. Skills like converting-mcps-to-skills, defragmenting-memory, and dispatching-coding-agents are directly applicable to Hermes workflows. The acquiring-skills skill installs from Hermes, ClawHub, and GitHub — creating a two-way skill ecosystem bridge.


Installation

# === Agenthood (14-agent team) ===
npx skills add fworks-tech/agenthood

# Or install individual agents:
npx skills add fworks-tech/agenthood --skill the-reviewer
npx skills add fworks-tech/agenthood --skill the-architect
npx skills add fworks-tech/agenthood --skill the-auditor

# === Letta Code Agent Harness (24 skills) ===
npm install -g @letta-ai/letta-code
letta skills install https://github.com/letta-ai/letta-code

Impact Assessment

Agenthood fills the multi-agent architecture gap in our marketplace. With 14 clearly-defined roles, it provides a reference implementation for Hermes agent teams. The skills.sh compatibility means Hermes agents can directly load individual roles (e.g., the-reviewer for PR review workflows, the-auditor for security scanning).

Letta Code completes our documentation of one of the most sophisticated agent harnesses in the ecosystem. The memory architecture (MemFS with git backing, memory defragmentation), MCP-to-skills conversion pipeline, and self-improving agent loops are patterns Hermes can learn from. The bidirectional skill bridge (acquiring-skills installs from Hermes) creates ecosystem interoperability.


June 26 Morning Update | Marketplace Home