“Developers looking for skills or plugins that plan, review, harden, or simplify code changes inside Hermes Agent or Claude Code.”
Structured methodologies and subagent swarms that turn ad-hoc coding sessions into repeatable, checkpointed workflows — spec-first planning, adversarial code review, and multi-agent simplification passes.
Updated 2026-08-20 · tested against Hermes v0.20.4
The most-starred agentic skills framework in the catalog by two orders of magnitude. If you install exactly one dev-workflow plugin, this is the default.
hermes plugins install obra/superpowers --enable
works withHermes Agent CLI (plugin); also ships as a Claude Code plugin via the official marketplace.
Built specifically for Hermes' own Skills Hub syntax rather than ported from Claude Code — teaches the agent to write its own analytical prompts via prism-scan plus seven proven analytical lenses.
hermes skills tap add Cranot/super-hermes
works withHermes Agent native (Skills Hub tap); manual git-clone install also documented for other harnesses.
Turns your own repo, docs, and developer discourse into agentskills.io-compliant skills with file/line provenance citations — for teams standardizing internal skills, not consuming someone else's.
npx bmad-module-skill-forge install
works withNode.js >= 22, Python >= 3.10, uv required; IDE-agnostic scaffolding tool, not Hermes-specific.
Three parallel subagents attack code from different angles, then verify findings via sandboxed proof-of-concept execution before reporting VERIFIED or REFUTED — the most rigorous read-only code-review skill in this list.
Clone the repo and copy it into your harness's skill directory (pip install . provides its 9 console scripts); no Skills Hub listing yet.
works withHarness-agnostic Agent Skill, Python-backed; works with Hermes, Claude Code, and any agentskills.io-compatible client.
Three read-only subagents (Hygiene, Clarity, Correctness) with a SAFE/CAREFUL/RISKY tiered approval gate before anything is applied — a safer default than a single autonomous simplification pass.
git clone https://github.com/Sahil-SS9/hermes-simplify-swarm.git, then copy SKILL.md and references/ into your harness's skill directory as simplify-swarm/.
works withHarness-agnostic (copy-in SKILL.md); explicitly built for Hermes Agent, also works in Claude Code.
These are curated by hand, not scraped or ranked by an algorithm. Every entry is a project already accepted into the Atlas catalog; the install command is copied out of that project's own README (not inferred), the compatibility line reflects what the README actually claims to support, and the verdict says what this pick does that its neighbours don't. Star counts come live from the GitHub API on each build. Updated 2026-08-20 · tested against Hermes v0.20.4.
obra/superpowers. The most-starred agentic skills framework in the catalog by two orders of magnitude. If you install exactly one dev-workflow plugin, this is the default. Install it with: hermes plugins install obra/superpowers --enable
From the Hermes CLI. Running hermes skills install owner/repo/skills/name installs a single skill; hermes skills tap add owner/repo registers a whole repository as a tap, so every skill inside it becomes available. Plugins use hermes plugins install owner/repo --enable. Skills that follow the agentskills.io spec but aren't published to the Skills Hub can be added with npx skills add owner/repo, or cloned straight into your skills directory. Every entry on this page carries the exact command from its own README.
Updated 2026-08-20 · tested against Hermes v0.20.4. The picks are curated by hand — install commands are read out of each project's README rather than generated, and star counts are refreshed from the GitHub API on every build.