Super Hermes
Cranot/super-hermes
Skills that teach Hermes Agent to write its own analytical prompts and report what it missed
Super Hermes is a skill pack that has Hermes Agent write its own analysis instructions before a complex task, then report both what it found and what it could not examine. It adds slash commands such as /prism-scan and /prism-full.
What Super Hermes does
Super Hermes teaches Hermes Agent to write its own thinking instructions before a complex task. The agent generates a cognitive prism tailored to the problem, executes it, and reports its findings together with a constraint report that names what the analysis did not examine and which follow-up command to run. The README contrasts a vanilla answer that names a pattern with a prism answer that derives structural trade-offs, and it validates the approach on Starlette, Click and Tenacity source files, with outputs in the examples folder.
Five skills are provided. /prism-scan generates and runs a lens for an artifact, /prism-full runs a multi-pass pipeline with adversarial self-correction, /prism-3way analyzes WHERE, WHEN and WHY on code, business, strategy or text, /prism-discover maps possible analysis domains, and /prism-reflect adds a meta-analysis of what the first pass concealed. Seven proven prisms, such as error_resilience.md and l12.md for conservation laws, ship inside the prism-scan skill and are also copied to ~/.hermes/prisms/.
Key features
- /prism-scan writes and executes a custom analytical lens for an artifact
- /prism-full multi-pass pipeline with adversarial self-correction
- /prism-3way analysis of WHERE, WHEN and WHY on code, business, strategy or text
- /prism-discover to map possible analysis domains
- /prism-reflect meta-analysis with a constraint transparency report
- Seven proven prisms stored in skills/prism-scan/references and ~/.hermes/prisms/
When to use it
- Review a module for structural trade-offs rather than a surface checklist
- Find bugs with line references and severity in an unfamiliar file
- Get a list of what an analysis did not cover and what to run next
Who it is for: Hermes Agent users who do code review, architecture analysis or strategy work and want deeper, self-critical analysis.
How it fits with Hermes Agent
It is built for Hermes Agent: the skills run inside Hermes, and both installers copy the prisms to ~/.hermes/prisms/. The README says the prisms can also be used with any tool that accepts system prompts.
FAQ
What is Super Hermes?
Super Hermes is a set of skills for Hermes Agent. Before a complex task the agent writes its own analytical prompt, runs it, and reports what it found and what it could not see.
Does Super Hermes work with Hermes Agent?
Yes. It is written for Hermes Agent and adds slash commands such as /prism-scan, /prism-full, /prism-3way, /prism-discover and /prism-reflect. The installers also place the prism files in ~/.hermes/prisms/.
Is Super Hermes free and open source?
Yes. The repository is released under the MIT license.
Similar skills for Hermes Agent
All skillsMeta-skill and plugin that spots repeated Hermes workflows and writes them as SKILL.md files
alchaincyf Elon Musk SkillAgent skill that applies Elon Musk's mental models and decision heuristics, built from public sources
smixs Visual SkillsTwo agent skills that write cinematic AI video and image prompts with model-specific syntax
cosmicstack-labs Mercury SkillsOpen library of 132 SKILL.md playbooks with install paths for Mercury, Hermes, Claude Code and Codex
geekjourneyx Claude Design CardAgent skill that turns text or URLs into publishable design cards in 14 formats
HuangYuChuh ComfyUI Skills for OpenClawTurns ComfyUI workflows into callable agent skills via a CLI, for Hermes Agent, OpenClaw and Codex
Related guides: What is Hermes Agent? · SOUL.md for Hermes Agent: what it is and how to write one