obra/superpowers
A composable skill and workflow bundle, installable as a Hermes plugin, that drives coding agents through brainstorming, specification, planning, red-green-refactor TDD and subagent-driven implementation.
Awesome Hermes Plugins › Commands & Workflows
Bergschloss/Refine-Cycle-for-Hermes-Agent
Refine Cycle is a Hermes Agent plugin that targets mistakes an agent keeps repeating across different sessions rather than within a single conversation. It fingerprints errors into normalized, comparable shapes (collapsing volatile details like ids, paths, ports and timestamps) and counts recurrence both within and across sessions; once a problem has appeared in at least two sessions or five times, it proposes and, after a further check, applies the smallest useful fix - a short memory entry, a new or improved skill, or a focused note - then watches later sessions and grades whether that fix is working, unused, unreliable, or too new to judge. It is adapted from the /refine concept in Prime Intellect's Prime Agent (Continual Harness). The plugin is explicitly bounded: at most three changes per day, every change journaled with a one-command rollback when safe, the base system prompt never touched, and only agent-created skills and memory entries ever edited (built-in, pinned and hub-installed skills are off-limits). Its installer also disables Hermes's write_approval queue for memory and skills, since lessons cannot land while writes wait for manual approval, and it backs up the config file first; a plugin-only mode skips that and a related memory-limit change. A published pre-registered experiment (133 probes, four arms) found the agent completed the graded behavior 66 out of 133 times with a lesson in memory versus 28 out of 133 with none, with placebo lessons scoring similarly to the empty-memory baseline, though the authors note real caveats: several lessons pass mainly by naming the trigger condition explicitly, and results differed in sign on an earlier pilot route. All three installer effects are reversible with a rollback command. It targets teams running Hermes Agent long enough to accumulate recurring, session-spanning failures and willing to let a bounded, auditable process manage the fix.
https://github.com/Bergschloss/Refine-Cycle-for-Hermes-Agent
Also in Commands & Workflows
OthmanAdi/planning-with-files
This repository provides a persistent, file-based planning skill for AI coding agents, ensuring plans survive context loss, restarts, and offering a completion gate for long-running tasks.
rlaope/oh-my-hermes
Operating layer for Hermes Agent that adds eleven routed workflows covering planning, research, creation, coding handoffs, operations and project memory with explicit evidence boundaries.
Lifecycle-Innovations-Limited/claude-ops
A business operations plugin that ports to Hermes Agent, gathering infrastructure health, CI/CD, inbox, pull requests, sprint state and revenue into one briefing plus autonomous agents.
lingjiuu/hermes-dynamic-workflows
This plugin introduces dynamic workflows to Hermes Agent, allowing models to write, execute, and orchestrate sandboxed Python scripts for complex multi-agent tasks.
xuyang-liu16/hermes-code-bridge
This Hermes Agent plugin acts as a control plane, enabling coordination and routing of tasks to local coding agents like Codex, Kimi Code, Claude Code, OpenCode, and Gemini CLI.
mrrkrieg/loopgraph
Native Hermes plugin and governance layer that routes verified company events into department-owned business loops, validating each routing decision, controlling execution and recording the outcome.
MahdiHedhli/HermesUltraCode
Hermes Agent plugin that gates every subagent dispatch behind a review by a model from a different lab, allowing only tightening or blocking, and records each decision in an immutable audit trail.