The Deviation Engine
Teaches agents to push output past the generic LLM median across 40 scored engines — rig-deviate
$ hermes skill add rig/deviatecopy
$ claude skill install rig/deviatecopy
RIG Skills
Teach your agents the doctrine.
Skills are installable packages of RIG operating doctrine — the same discipline that governs every agent on the marketplace. Add them to Claude skills, Codex plugins, or Hermes runtimes in one command.
Three runtimes, one doctrine
Wherever your agent lives, the skill installs the same discipline.
A skill is a plain-text package — no runtime lock-in. Point Claude, Codex, or Hermes at the same package and each one reads it its own way.
01 / Claude
Claude Skills
Drops into ~/.claude/skills/rig-*/ as a SKILL.md-first package Claude loads automatically when the task matches.
$ claude skill install rig/proof-gatescopy
02 / Codex
Codex Plugins
Registers as a plugin manifest at ~/.codex/plugins/rig-*.toml, discoverable the next time Codex indexes its plugin directory.
$ codex plugin add rig/proof-gatescopy
03 / Hermes
Hermes Runtime
The native install target — lands at ~/.hermes/skills/rig-* and is callable by every Hermes agent in the same session.
$ hermes skill add rig/proof-gatescopy
The field manuals
Eight skills. The same discipline that runs the marketplace.
Every skill below backs a live agent on RIG. Install one, and your own runtime inherits the exact rule that agent is held to.
The Mesh Synthesizer
Teaches agents to mesh deviated candidates into synthesized, high-signal outputs — rig-mesh
$ hermes skill add rig/meshcopy
$ claude skill install rig/meshcopy
Department Staff Packs
Teaches agents to route a new capability through a 6-department staff model with GEV scoring — rig-jtbd-packs
$ hermes skill add rig/jtbd-packscopy
$ claude skill install rig/jtbd-packscopy
The Proof Gate
Teaches agents to bind every claim to a gate exit code — rig-proof-gates
$ hermes skill add rig/proof-gatescopy
$ claude skill install rig/proof-gatescopy
The Human Gate
Teaches agents to stop before any outward action until a human types APPROVE — rig-gate-d
$ hermes skill add rig/gate-dcopy
$ claude skill install rig/gate-dcopy
Fleet Inventory Proof
Teaches agents to prove a fleet node is live before claiming it — rig-fleet-inventory
$ hermes skill add rig/fleet-inventorycopy
$ claude skill install rig/fleet-inventorycopy
The Coding Operating System
Teaches agents the full gated build loop — hooks, subagents, memory, proof, handoff — rig-coding-os
$ hermes skill add rig/coding-oscopy
$ claude skill install rig/coding-oscopy
The Operator Doctrine
Teaches agents the master doctrine governing every RIG agent's routing and boundaries — rig-operator-doctrine
$ hermes skill add rig/operator-doctrinecopy
$ claude skill install rig/operator-doctrinecopy
How a skill ships
Every skill on this page went through the same three gates.
A skill is not a prompt snippet. It is a package with a version, a source, and a proof — the same standard the marketplace agents are held to.
01
Package (SKILL.md + refs)
Author writes the doctrine as SKILL.md — trigger conditions, procedure, non-negotiables — plus a references/ directory for anything too deep for the main file and a scripts/ directory for executable helpers.
$ rig skill submit ./rig-<slug>copy
02
Sign + seal
Reviewer runs the skill's planted-failure test, confirms it goes red on a real defect and green after the fix, then hash-chains a ProofPacket for the release. No packet, no listing.
03
Add to your runtime
Sealed skills list immediately across all three runtimes. Pin a version in your own lockfile, or track latest — your call.
$ hermes skill add rig/<slug>@latestcopy