Skip to content
Back to search

agentsys

MarketplaceWorks with Claude CodeGitHub

11 specialized plugins for AI workflow automation - GitHub issue queue for native agents, slop detection, negative behavior memory, skill and system prompt curation, topic research, agent config linting, workflow pattern learning, Zig language support, Mojo language support, and Ada/SPARK language support

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today
Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/agent-sh-agentsys/badge.svg)](/listings/agent-sh-agentsys)

HTML

<a href="/listings/agent-sh-agentsys"><img src="/api/listings/agent-sh-agentsys/badge.svg" alt="findsafeskills safety rating" /></a>

Install

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

Claude Code (Recommended way)

/plugin marketplace add agent-sh/agentsys
/plugin install foreman@agentsys

All Platforms (npm)

npm install -g agentsys && agentsys

Interactive installer for Claude Code, OpenCode, Codex CLI, Cursor, and Kiro.

# Non-interactive install
agentsys --tool claude              # Single tool
agentsys --tool cursor              # Cursor (~/.cursor: skills, commands, rules)
agentsys --tool kiro                # Kiro (~/.kiro: prompts, skills, agents)
agentsys --tools "claude,opencode"  # Multiple tools
agentsys --development              # Dev mode (bypasses marketplace)
agentsys install deslop             # One plugin, or plugin:component for one agent, skill or command

[Full installation guide →](./docs/INSTALLATION.md)

---

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

.claude-plugin/marketplace.json · Manifest · 176 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

  1. {
  2. "name": "agentsys",
  3. "description": "11 specialized plugins for AI workflow automation - GitHub issue queue for native agents, slop detection, negative behavior memory, skill and system prompt curation, topic research, agent config linting, workflow pattern learning, Zig language support, Mojo language support, and Ada/SPARK language support",
  4. "version": "6.1.1",
  5. "owner": {
  6. "name": "Avi Fenesh",
  7. "url": "https://github.com/avifenesh"
  8. },
  9. "repository": "https://github.com/agent-sh/agentsys",
  10. "keywords": [
  11. "ai",
  12. "llm",
  13. "agents",
  14. "agentic",
  15. "claude-code",
  16. "opencode",
  17. "codex",
  18. "mcp",
  19. "automation",
  20. "workflow",
  21. "code-review",
  22. "multi-agent"
  23. ],
  24. "plugins": [
  25. {
  26. "name": "foreman",
  27. "source": {
  28. "source": "url",
  29. "url": "https://github.com/agent-sh/foreman.git",
  30. "commit": "01a4964590072a747658741c26a3d306129b236d",
  31. "sha": "01a4964590072a747658741c26a3d306129b236d"
  32. },
  33. "description": "GitHub issues as a task queue for native agents: pick, claim, worktree workers, Revuto findings, land through the repo's own land command",
  34. "version": "0.1.5",
  35. "category": "productivity",
  36. "homepage": "https://github.com/agent-sh/foreman"
  37. },
  38. {
  39. "name": "deslop",
  40. "source": {
  41. "source": "url",
  42. "url": "https://github.com/agent-sh/deslop.git",
  43. "commit": "5b67bf191c135d8afc3ef3448a1a503078902a1c",
  44. "sha": "5b67bf191c135d8afc3ef3448a1a503078902a1c"
  45. },
  46. "description": "3-phase AI slop detection: regex patterns (HIGH), multi-pass analyzers (MEDIUM), CLI tools (LOW)",
  47. "version": "1.3.0",
  48. "category": "development",
  49. "homepage": "https://github.com/agent-sh/deslop"
  50. },
  51. {
  52. "name": "skill-curator",
  53. "source": {
  54. "source": "url",
  55. "url": "https://github.com/agent-sh/skill-curator.git",
  56. "ref": "v1.2.0",
  57. "commit": "62f7a2682b3d04830abfbd72c163dada97e875de",
  58. "sha": "62f7a2682b3d04830abfbd72c163dada97e875de"
  59. },
  60. "description": "Canonical guidance for writing, reviewing, and maintaining production-grade SKILL.md files across agent tools",
  61. "version": "1.2.0",
  62. "category": "development",
  63. "homepage": "https://github.com/agent-sh/skill-curator"
  64. },
  65. {
  66. "name": "system-prompt-curator",
  67. "source": {
  68. "source": "url",
  69. "url": "https://github.com/agent-sh/system-prompt-curator.git",
  70. "ref": "v2.1.0",
  71. "commit": "b51039f4dbc53229c3cfd310293ee34bec4a90e7",
  72. "sha": "b51039f4dbc53229c3cfd310293ee34bec4a90e7"
  73. },
  74. "description": "Production-grade system prompt curation for autonomous coding agents and multi-agent orchestration",
  75. "version": "2.1.0",
  76. "category": "development",
  77. "homepage": "https://github.com/agent-sh/system-prompt-curator"
  78. },
  79. {
  80. "name": "learn",
  81. "source": {
  82. "source": "url",
  83. "url": "https://github.com/agent-sh/learn.git",
  84. "commit": "37e674dce2be988b1709055d08793c16c99987b9",
  85. "sha": "37e674dce2be988b1709055d08793c16c99987b9"
  86. },
  87. "description": "Research topics online and create comprehensive learning guides with RAG-optimized indexes",
  88. "version": "1.3.0",
  89. "category": "productivity",
  90. "homepage": "https://github.com/agent-sh/learn"
  91. },
  92. {
  93. "name": "banthis",
  94. "source": {
  95. "source": "url",
  96. "url": "https://github.com/agent-sh/banthis.git",
  97. "ref": "v0.6.0",
  98. "commit": "700cb0eea3f9de83d22724d6de38b01f7971cf1f",
  99. "sha": "700cb0eea3f9de83d22724d6de38b01f7971cf1f"
  100. },
  101. "description": "Durable negative behavior memory: persist user-defined banned agent behaviors into CLAUDE.md or AGENTS.md",
  102. "version": "0.6.0",
  103. "category": "productivity",
  104. "homepage": "https://github.com/agent-sh/banthis"
  105. },
  106. {
  107. "name": "agnix",
  108. "source": {
  109. "source": "git-subdir",
  110. "url": "https://github.com/agent-sh/agnix.git",
  111. "path": "plugin",
  112. "commit": "2c0e4efede3181599eca37619e8067ae2d941c00",
  113. "sha": "2c0e4efede3181599eca37619e8067ae2d941c00"
  114. },
  115. "description": "Lint agent configuration files (SKILL.md, CLAUDE.md, hooks, MCP) against 455 rules across 10+ AI tools",
  116. "version": "0.57.0",
  117. "category": "development",
  118. "homepage": "https://github.com/agent-sh/agnix"
  119. },
  120. {
  121. "name": "skillers",
  122. "source": {
  123. "source": "url",
  124. "url": "https://github.com/agent-sh/skillers.git",
  125. "commit": "682e3355bdede71bfbdc3cc61e8f3ddf461a3c56",
  126. "sha": "682e3355bdede71bfbdc3cc61e8f3ddf461a3c56",
  127. "ref": "v0.3.0"
  128. },
  129. "description": "Learn from workflow patterns across sessions and suggest skills, hooks, and agents to automate repetitive work",
  130. "version": "0.3.0",
  131. "category": "productivity",
  132. "homepage": "https://github.com/agent-sh/skillers"
  133. },
  134. {
  135. "name": "zig-lsp",
  136. "source": {
  137. "source": "url",
  138. "url": "https://github.com/agent-sh/zig-lsp.git",
  139. "commit": "ecf32677d7bb3cc8e28112bb3c70dd7809a3aea7",
  140. "sha": "ecf32677d7bb3cc8e28112bb3c70dd7809a3aea7"
  141. },
  142. "description": "Zig language server for Claude Code via ZLS - automatic diagnostics after every edit, jump-to-definition, find-references, and hover. Requires zls in PATH",
  143. "version": "0.1.1",
  144. "category": "development",
  145. "homepage": "https://github.com/agent-sh/zig-lsp"
  146. },
  147. {
  148. "name": "mojo",
  149. "source": {
  150. "source": "url",
  151. "url": "https://github.com/agent-sh/mojo.git",
  152. "ref": "v0.3.0",
  153. "commit": "f6bac33ff5fa61f0fcf87349a5213758cb1f4dc4",
  154. "sha": "f6bac33ff5fa61f0fcf87349a5213758cb1f4dc4"
  155. },
  156. "description": "Teach agents to write idiomatic, correct, performant, current Mojo (v1.0.0b1) - syntax, ownership and CPU/memory optimization (copies, SIMD, vectorize/parallelize), GPU kernels (DeviceContext/LayoutTensor), and Mojo/Python interop; prevents stale pre-2025 syntax",
  157. "version": "0.3.0",
  158. "category": "development",
  159. "homepage": "https://github.com/agent-sh/mojo"
  160. },
  161. {
  162. "name": "ada-spark",
  163. "source": {
  164. "source": "url",
  165. "url": "https://github.com/agent-sh/ada-spark.git",
  166. "ref": "v0.2.0",
  167. "commit": "cecb406acd326db38efac7fd475e846fd3541d24",
  168. "sha": "cecb406acd326db38efac7fd475e846fd3541d24"
  169. },
  170. "description": "Teach agents to write idiomatic, correct, current Ada and SPARK (Ada 2022) - contracts (aspects, Pre'Class), the Alire ecosystem, SPARK proof (AoRTE, assurance levels, ownership/borrow), embedded (Ravenscar/Jorvik), and GNAT/GNAT SAS tooling; blocks stale pre-2022 advice (GNAT Community, pragma contracts, CodePeer)",
  171. "version": "0.2.0",
  172. "category": "development",
  173. "homepage": "https://github.com/agent-sh/ada-spark"
  174. }
  175. ]
  176. }
README.md · README · 698 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

  1. <p align="center">
  2. <img src="site/assets/logo.png" alt="AgentSys" width="120">
  3. </p>
  4. <h1 align="center">AgentSys</h1>
  5. <p align="center">
  6. <strong>A modular runtime and orchestration system for AI agents.</strong>
  7. </p>
  8. <p align="center">
  9. <a href="https://www.npmjs.com/package/agentsys"><img src="https://img.shields.io/npm/v/agentsys.svg" alt="npm version"></a>
  10. <a href="https://www.npmjs.com/package/agentsys"><img src="https://img.shields.io/npm/dm/agentsys.svg" alt="npm downloads"></a>
  11. <a href="https://github.com/agent-sh/agentsys/actions/workflows/ci.yml"><img src="https://github.com/agent-sh/agentsys/actions/workflows/ci.yml/badge.svg" alt="CI"></a>
  12. <a href="https://github.com/agent-sh/agentsys/stargazers"><img src="https://img.shields.io/github/stars/agent-sh/agentsys.svg" alt="GitHub stars"></a>
  13. <a href="https://opensource.org/licenses/MIT"><img src="https://img.shields.io/badge/License-MIT-yellow.svg" alt="License: MIT"></a>
  14. <a href="https://agent-sh.github.io/agentsys/"><img src="https://img.shields.io/badge/Website-AgentSys-blue?style=flat&logo=github" alt="Website"></a>
  15. <a href="https://github.com/hesreallyhim/awesome-claude-code"><img src="https://awesome.re/mentioned-badge.svg" alt="Mentioned in Awesome Claude Code"></a>
  16. </p>
  17. <p align="center">
  18. <b>11 plugins · 5 agents · 14 skills (across all repos) · 13k lines of lib code · tests on Linux and Windows · 5 platforms</b><br>
  19. <em>Plugins distributed as standalone repos under <a href="https://github.com/agent-sh">agent-sh</a> org - agentsys is the marketplace &amp; installer</em>
  20. </p>
  21. > Running open models for your company? [**Tiyuvta**](https://tiyuvta.ai/services/) helps with model choice, deployment and optimization, and fine-tuning on your hardware or cloud account.
  22. <p align="center">
  23. <a href="https://github.com/sponsors/avifenesh"><img src="https://img.shields.io/badge/Sponsor-%E2%9D%A4-db61a2?style=for-the-badge&logo=githubsponsors&logoColor=white" alt="Sponsor"></a>
  24. </p>
  25. <p align="center"><b>Built and maintained with a lot of agent tokens and hours. If it saves you time, <a href="https://github.com/sponsors/avifenesh">sponsoring</a> helps keep it going.</b></p>
  26. <p align="center">
  27. <a href="#commands">Commands</a> · <a href="#installation">Installation</a> · <a href="https://agent-sh.github.io/agentsys/">Website</a> · <a href="https://github.com/agent-sh/agentsys/discussions">Discussions</a>
  28. </p>
  29. <p align="center">
  30. <b>Built for Claude Code · Codex CLI · OpenCode · Cursor · Kiro</b>
  31. </p>
  32. <p align="center"><em>New skills, agents, and integrations ship constantly. Follow for real-time updates:</em></p>
  33. <p align="center">
  34. <a href="https://x.com/avi_fenesh"><img src="https://img.shields.io/badge/Follow-@avi__fenesh-1DA1F2?style=for-the-badge&logo=x&logoColor=white" alt="Follow on X"></a>
  35. </p>
  36. ---
  37. AI models can write code. That's not the hard part anymore. The hard part is everything around it - branch management, code review, artifact cleanup, CI, PR comments. **AgentSys is the runtime that orchestrates agents to handle all of it** - structured pipelines, gated phases, specialized agents, and persistent state that survives session boundaries.
  38. ---
  39. > Building custom skills, agents, hooks, or MCP tools? [agnix](https://github.com/agent-sh/agnix) is the CLI + LSP linter that catches config errors before they fail silently - real-time IDE validation, auto suggestions, auto-fix, and 757 rules for Claude Code, Codex, OpenCode, Cursor, Kiro, Copilot, Gemini CLI, Cline, Windsurf, Roo Code, Amp, and more.
  40. ## What's New in 6.1.1
  41. - OpenCode, Codex, Cursor and Kiro installs get the plugins that ship no `.claude-plugin/plugin.json` at their pin. The installer fetched such a plugin and installed nothing from it. It now makes one from the marketplace entry, as Claude Code does.
  42. - agnix installs its `/agnix` command, agent and skill on every platform. Claude Code installed only its skill before, and the other platforms nothing. The marketplace entry is now a `git-subdir` source for the plugin's `plugin/` folder.
  43. - The installer fetches a `git-subdir` source as that folder, pinned by `sha` or `commit`.
  44. - Contributor docs: every change reaches main through a PR, and a release commit merges through a PR with only the tag pushed.
  45. ## What This Is
  46. An agent orchestration system - 11 plugins, 5 agents, and 14 skills that compose into structured pipelines for software development. Each plugin lives in its own standalone repo under the [agent-sh](https://github.com/agent-sh) org. agentsys is the marketplace and installer that ties them together.
  47. Each agent has a single responsibility, a specific model assignment, and defined inputs/outputs. Pipelines enforce phase gates so agents can't skip steps. State persists across sessions so work survives interruptions.
  48. The system runs on Claude Code, OpenCode, Codex CLI, Cursor, and Kiro. Install via the marketplace or the npm installer, and the plugins are fetched automatically from their repos.
  49. ---
  50. ## The Approach
  51. **Code does code work. AI does AI work.**
  52. - **Detection**: regex, AST analysis, static analysis - fast, deterministic, no tokens wasted
  53. - **Judgment**: LLM calls for synthesis, planning, review - where reasoning matters
  54. - **Result**: certainty-graded findings throughout
  55. **Certainty levels exist because not all findings are equal:**
  56. | Level | Meaning | Action |
  57. |-------|---------|--------|
  58. | HIGH | Definitely a problem | Safe to auto-fix |
  59. | MEDIUM | Probably a problem | Needs context |
  60. | LOW | Might be a problem | Needs human judgment |
  61. This came from testing on 1,000+ repositories.
  62. ---
  63. ## Commands
  64. | Command | What it does |
  65. |---------|--------------|
  66. | [`/foreman`](#foreman) | Work the GitHub issue queue with native subagents, from pick to land |
  67. | [`/banthis`](#banthis) | Durable negative memory: persist banned agent behaviors |
  68. | [`/agnix`](#agnix) | Lint agent configurations (757 rules) |
  69. | [`/deslop`](#deslop) | Clean AI slop patterns |
  70. | [`/learn`](#learn) | Research topics, create learning guides |
  71. | [`/skillers`](#skillers) | Workflow pattern learning and automation |
  72. | [`/skill-curator`](#skill-curator) | Create and improve reliable SKILL.md files |
  73. | [`/system-prompt-curator`](#system-prompt-curator) | Create and improve autonomous agent system prompts |
  74. Each command works standalone. Together, they compose into end-to-end pipelines.
  75. ---
  76. ## Skills
  77. 14 skills included across the plugins:
  78. | Category | Skills |
  79. |----------|--------|
  80. | **Workflow** | `foreman` |
  81. | **Message Queues** | `glide-mq-migrate-bee`, `glide-mq-migrate-bullmq`, `glide-mq` |
  82. | **Enhancement** | `skill-curator`, `system-prompt-curator` |
  83. | **Cleanup** | `deslop` |
  84. | **AI Collaboration** | `learn`, `recommend`, `skillers-compact` |
  85. | **Memory** | `banthis` |
  86. | **Linting** | `agnix` |
  87. | **Languages** | `ada-spark`, `mojo` |
  88. **External skill plugins** (standalone repos, installed separately):
  89. | Category | Skills | Plugin |
  90. |----------|--------|--------|
  91. | **Message Queues** | `glide-mq`, `glide-mq-migrate-bullmq`, `glide-mq-migrate-bee` | [agent-sh/glidemq](https://github.com/agent-sh/glidemq) |
  92. Skills are the reusable implementation units. Agents invoke skills; commands orchestrate agents. When you install a plugin, its skills become available to all agents in that session.
  93. ---
  94. ## Quick Navigation
  95. | Section | What's there |
  96. |---------|--------------|
  97. | [The Approach](#the-approach) | Why it's built this way |
  98. | [Commands](#commands) | Overview of every command |
  99. | [Skills](#skills) | 14 skills across plugins |
  100. | [Skill-Only Plugins](#skill-only-plugins) | glide-mq and other non-command plugins |
  101. | [Command Details](#command-details) | Deep dive into each command |
  102. | [How Commands Work Together](#how-commands-work-together) | Standalone vs integrated |
  103. | [Design Philosophy](#design-philosophy) | The thinking behind the architecture |
  104. | [Installation](#installation) | Get started |
  105. | [Research & Testing](#research--testing) | What went into building this |
  106. | [Documentation](#documentation) | Links to detailed docs |
  107. ---
  108. ## Skill-Only Plugins
  109. Plugins without a `/` command. The marketplace ones install with agentsys; glide-mq installs from its own repo.
  110. | Plugin | What it does |
  111. |--------|--------------|
  112. | [mojo](https://github.com/agent-sh/mojo) | `mojo` skill: current Mojo syntax, ownership, SIMD and GPU kernels, Python interop |
  113. | [ada-spark](https://github.com/agent-sh/ada-spark) | `ada-spark` skill: Ada 2022 and SPARK contracts, proof, Alire and GNAT tooling |
  114. | [zig-lsp](https://github.com/agent-sh/zig-lsp) | Zig language server for Claude Code via ZLS (no skill). Needs `zls` on `PATH` |
  115. ### glide-mq
  116. Build message queues, background jobs, and workflow orchestration with [glide-mq](https://github.com/avifenesh/glide-mq) - high-performance Node.js queue on Valkey/Redis. Not in the agentsys marketplace: install its skills from [agent-sh/glidemq](https://github.com/agent-sh/glidemq).
  117. | Skill | What it does |
  118. |-------|--------------|
  119. | `glide-mq` | Greenfield queue development - queues, workers, ordering, rate limiting, flows, broadcast, step jobs |
  120. | `glide-mq-migrate-bullmq` | Migrate from BullMQ to glide-mq - API mapping, breaking changes, feature comparison |
  121. | `glide-mq-migrate-bee` | Migrate from Bee-Queue to glide-mq - API mapping, pattern conversion |
  122. Key features: per-key ordering, group concurrency, runtime group rate limiting (`job.rateLimitGroup()`), token bucket, DAG workflows, broadcast pub/sub, step jobs, deduplication, serverless producers.
  123. [Skill plugin →](https://github.com/agent-sh/glidemq) | [glide-mq docs →](https://avifenesh.github.io/glide-mq.dev/) | [npm →](https://www.npmjs.com/package/glide-mq)
  124. ---
  125. ## Command Details
  126. ### /foreman
  127. **Purpose:** Work a repo's GitHub issues as a task queue. One main session picks the next ready issue, claims it, and hands it to a worktree subagent. Revuto reviews the PR, and foreman lands it at the exact head it checked. It replaces next-task and tower-crane.
  128. There is no daemon and no state store. Labels, assignees and comments on GitHub are the state, and every command reads GitHub live, so several sessions can share one queue.
  129. **The loop:**
  130. 1. **Capacity, pick, claim** - checks load, CPU pressure and memory, takes the next ready issue (urgent first, then oldest, open dependencies skipped) and claims it
  131. 2. **Brief and spawn** - one worktree subagent per issue from its brief, in parallel up to `agents_max`
  132. 3. **Review** - reads Revuto's findings; every real finding gets fixed, a wrong one gets a reply. When Revuto is capped or down, the session reviews the diff itself and records it
  133. 4. **Land** - every check green at the head, no unaddressed finding, the repo's own land command, then a squash merge pinned to that head and the issue closed
  134. ```bash
  135. /foreman # Run the queue until nothing is ready
  136. /foreman status # Counts by label, claims and capacity
  137. /foreman stats # Revuto caps, self-reviews, manual review requests
  138. /foreman 123 # Work issue #123 only, from claim to land
  139. ```
  140. **Config:** optional `.foreman.json` in the repo root: label names, `agents_max`, load and memory holds, the check and land commands, required checks, Revuto round cap, forbidden words for posted comments.
  141. **Requires:** Node 22+, `gh` (logged in), `git` and `waitfor`. `run-long` needs systemd user units (Linux).
  142. **Skill:** foreman (no agents: workers are native worktree subagents)
  143. [Full workflow documentation →](./docs/workflows/FOREMAN.md) · [Repo →](https://github.com/agent-sh/foreman)
  144. ---
  145. ### /banthis
  146. **Purpose:** Durable negative memory for repeated agent mistakes. Turn a user's "stop doing this" correction into a persistent rule in `CLAUDE.md` or `AGENTS.md`.
  147. **[banthis](https://github.com/agent-sh/banthis)** is a tiny standalone CLI plus skill. It maintains a managed banned-behaviors section, supports project or global targets, and includes an `init` meta-rule so agents learn when to invoke it automatically.
  148. **What it does:**
  149. | Command | Use |
  150. |---------|-----|
  151. | `banthis add "<title>" "<rule>"` | Add or update a banned behavior |
  152. | `banthis list` | List current bans |
  153. | `banthis show` | Print the managed section |
  154. | `banthis remove "<title>"` | Remove a ban |
  155. | `banthis init` | Install the meta-rule that teaches agents to call `banthis` |
  156. **Usage:**
  157. ```bash
  158. /banthis "stop ending with vague optional follow-up offers"
  159. banthis add "No vague endings" "Do not end with vague optional follow-up offers."
  160. banthis init --file AGENTS.md
  161. ```
  162. ---
  163. ### /agnix
  164. **Purpose:** Lint agent configurations before they break your workflow. The first dedicated linter for AI agent configs.
  165. **[agnix](https://github.com/agent-sh/agnix)** is a standalone open-source project that provides the validation engine. This plugin integrates it into your workflow.
  166. **The problem it solves:**
  167. Agent configurations are code. They affect behavior, security, and reliability. But unlike application code, they have no linting. You find out your SKILL.md is malformed when the agent fails. You discover your hooks have security issues when they're exploited. You realize your CLAUDE.md has conflicting rules when the AI behaves unexpectedly.
  168. agnix catches these issues before they cause problems.
  169. **What it validates:**
  170. | Category | What It Checks |
  171. |----------|----------------|
  172. | **Structure** | Required fields, valid YAML/JSON, proper frontmatter |
  173. | **Security** | Prompt injection vectors, overpermissive tools, exposed secrets |
  174. | **Consistency** | Conflicting rules, duplicate definitions, broken references |
  175. | **Best Practices** | Tool restrictions, model selection, trigger phrase quality |
  176. | **Cross-Platform** | Compatibility across Claude Code, Codex, OpenCode, Cursor, Kiro, Copilot, Gemini CLI, Cline, Windsurf, Roo Code, Amp, and more |
  177. **757 validation rules** derived from:
  178. - Official tool specifications (Claude Code, Codex CLI, OpenCode, Cursor, Kiro, GitHub Copilot, Gemini CLI, Cline, Windsurf, Roo Code, Amp, and more)
  179. - Research papers on agent reliability and prompt injection
  180. - Real-world testing across 500+ repositories
  181. - Community-reported issues and edge cases
  182. **Supported files:**
  183. | File Type | Examples |
  184. |-----------|----------|
  185. | Skills | `SKILL.md`, `*/SKILL.md` |
  186. | Memory | `CLAUDE.md`, `AGENTS.md`, `.github/CLAUDE.md` |
  187. | Hooks | `.claude/settings.json`, hooks configuration |
  188. | MCP | `*.mcp.json`, MCP server configs |
  189. | Cursor | `.cursor/rules/*.mdc`, `.cursorrules` |
  190. | Copilot | `.github/copilot-instructions.md` |
  191. | Kiro | `.kiro/steering/**/*.md`, `.kiro/agents/*.json`, `.kiro/hooks/*.kiro.hook`, `POWER.md` |
  192. | Windsurf | `.windsurf/rules/**/*.md`, `.windsurf/workflows/**/*.md`, `.windsurfrules` |
  193. | Roo Code | `.roo/rules/*.md`, `.roo/rules-{mode}/*.md`, `.roomodes`, `.rooignore`, `.roorules` |
  194. | Gemini CLI | `GEMINI.md`, `.gemini/settings.json`, `gemini-extension.json` |
  195. | OpenCode | `opencode.json` |
  196. | Amp | `.agents/checks/**/*.md`, `.amp/settings.json` |
  197. **CI/CD Integration:**
  198. agnix outputs SARIF format for GitHub Code Scanning. Add it to your workflow:
  199. ```yaml
  200. - name: Lint agent configs
  201. run: agnix --format sarif > results.sarif
  202. - uses: github/codeql-action/upload-sarif@v3
  203. with:
  204. sarif_file: results.sarif
  205. ```
  206. **Usage:**
  207. ```bash
  208. /agnix # Validate current project
  209. /agnix --fix # Auto-fix fixable issues
  210. /agnix --strict # Treat warnings as errors
  211. /agnix --target claude-code # Only Claude Code rules
  212. /agnix --format sarif # Output for GitHub Code Scanning
  213. ```
  214. **Agent:** agnix-agent (sonnet model)
  215. **External tool:** Requires [agnix CLI](https://github.com/agent-sh/agnix)
  216. ```bash
  217. npm install -g agnix # Install via npm
  218. # or
  219. cargo install agnix-cli # Install via Cargo
  220. # or
  221. brew tap agent-sh/agnix && brew install agnix # Install via Homebrew
  222. ```
  223. **Why use agnix:**
  224. - Catch config errors before they cause agent failures
  225. - Enforce security best practices across your team
  226. - Maintain consistency as your agent configs grow
  227. - Integrate validation into CI/CD pipelines
  228. - Support multiple AI tools from one linter
  229. ---
  230. ### /deslop
  231. **Purpose:** Finds AI slop - debug statements, placeholder text, verbose comments, TODOs - and removes it.
  232. **How detection works:**
  233. Three phases run in sequence:
  234. 1. **Phase 1: Regex Patterns** (HIGH certainty)
  235. - `console.log`, `print()`, `dbg!()`, `println!()`
  236. - `// TODO`, `// FIXME`, `// HACK`
  237. - Empty catch blocks, disabled linters
  238. - Hardcoded secrets (API keys, tokens)
  239. 2. **Phase 2: Multi-Pass Analyzers** (MEDIUM certainty)
  240. - Doc-to-code ratio (excessive comments)
  241. - Verbosity ratio (AI preambles)
  242. - Over-engineering patterns
  243. - Buzzword inflation
  244. - Dead code detection
  245. - Stub functions
  246. 3. **Phase 3: CLI Tools** (LOW certainty, optional)
  247. - jscpd, madge, escomplex (JS/TS)
  248. - pylint, radon (Python)
  249. - golangci-lint (Go)
  250. - clippy (Rust)
  251. **Languages supported:** JavaScript/TypeScript, Python, Rust, Go, Java
  252. **Usage:**
  253. ```bash
  254. /deslop # Report only (safe)
  255. /deslop apply # Fix HIGH certainty issues
  256. /deslop apply src/ # Fix issues in src/ only
  257. ```
  258. **Thoroughness levels:**
  259. - `quick` - Phase 1 only (fastest)
  260. - `normal` - Phase 1 + Phase 2 (default)
  261. - `deep` - All phases if tools available
  262. [Pattern reference →](./docs/reference/SLOP-PATTERNS.md)
  263. ---
  264. ### /learn
  265. **Purpose:** Research any topic online and create a comprehensive learning guide with RAG-optimized indexes.
  266. **What it does:**
  267. 1. **Progressive Discovery** - Uses funnel approach (broad → specific → deep) to find quality sources
  268. 2. **Quality Scoring** - Scores sources by authority, recency, depth, examples, uniqueness
  269. 3. **Just-In-Time Extraction** - Fetches only high-scoring sources to save tokens
  270. 4. **Synthesis** - Creates structured learning guide with examples and best practices
  271. 5. **RAG Index** - Updates the `agent-knowledge/AGENTS.md` master index for future lookups
  272. **Depth levels:**
  273. | Depth | Sources | Use Case |
  274. |-------|---------|----------|
  275. | brief | 10 | Quick overview |
  276. | medium | 20 | Default, balanced |
  277. | deep | 40 | Comprehensive |
  278. **Output structure:**
  279. ```
  280. agent-knowledge/
  281. AGENTS.md # Master index (updated each run)
  282. recursion.md # Topic-specific guide
  283. resources/
  284. recursion-sources.json # Source metadata with quality scores
  285. ```
  286. **Usage:**
  287. ```bash
  288. /learn recursion # Default (20 sources)
  289. /learn react hooks --depth=deep # Comprehensive (40 sources)
  290. /learn kubernetes --depth=brief # Quick overview (10 sources)
  291. ```
  292. A `CLAUDE.md` mirror of the index is written only with `--legacy-claude-index`.
  293. **Agent:** learn-agent (sonnet model)
  294. ---
  295. ### /skillers
  296. > Learn from your workflow patterns and suggest automations
  297. ```bash
  298. /skillers show # Display current config and knowledge stats
  299. /skillers compact # Analyze recent transcripts, extract patterns
  300. /skillers compact --days=14 # Analyze older transcripts
  301. /skillers recommend # Get automation suggestions from accumulated knowledge
  302. ```
  303. Reads your Claude Code, Codex and OpenCode transcripts, identifies recurring patterns (pain points, repeated workflows, wishes), clusters them into weighted themes, and suggests skills, hooks, or agents to automate them.
  304. No per-turn overhead - it reads transcripts your tools already save.
  305. **Agents:** skillers-compactor (sonnet), skillers-recommender (opus)
  306. **Skills:** skillers-compact, recommend
  307. ---
  308. ### /skill-curator
  309. > Create and improve reliable `SKILL.md` files
  310. ```bash
  311. /skill-curator "create a skill for reviewing background jobs"
  312. /skill-curator --improve path/to/SKILL.md --category review
  313. ```
  314. The skill curator focuses on trigger quality, clear scope, router patterns, concrete `Skip unless:` gates, token budgets, and agnix-ready structure across Claude Code, Codex, OpenCode, Cursor, Kiro, and similar tools.
  315. **Skill:** skill-curator
  316. ---
  317. ### /system-prompt-curator
  318. > Create and improve autonomous coding-agent system prompts
  319. ```bash
  320. /system-prompt-curator "GitHub issue resolver"
  321. /system-prompt-curator --improve path/to/prompt.md --for-orchestrator
  322. ```
  323. The system prompt curator rewrites prompts around task-matched identity, phased workflow, explicit tools, evidence-based completion criteria, and realistic error recovery examples. It separates prompt guidance from harness-level checks that belong in code.
  324. **Skill:** system-prompt-curator
  325. ---
  326. ## How Commands Work Together
  327. **Standalone use:**
  328. ```bash
  329. /deslop apply # Just clean up your code
  330. /agnix # Just lint your agent configs
  331. /foreman # Work the issue queue, one worktree agent per issue
  332. /learn react hooks # Just research a topic
  333. ```
  334. **Together:**
  335. [`/foreman`](#foreman) picks an issue and hands it to a worktree worker. The worker can run [`/deslop`](#deslop) on its diff and [`/agnix`](#agnix) on any agent config it touched before it reports back. foreman then reads Revuto's findings and lands the PR only when every check is green at the head it checked.
  336. ---
  337. ## Design Philosophy
  338. <details>
  339. <summary><strong>Architecture decisions and trade-offs</strong> (click to expand)</summary>
  340. ### The Actual Problem
  341. Frontier models write good code. That's solved. What's not solved:
  342. - **Context management** - Models forget what they're doing mid-session
  343. - **Compaction amnesia** - Long sessions get summarized, losing critical state
  344. - **Task drift** - Without structure, agents wander from the actual goal
  345. - **Skipped steps** - Agents skip reviews, tests, or cleanup when not enforced
  346. - **Token waste** - Using LLM calls for work that static analysis can do faster
  347. - **Babysitting** - Manually orchestrating each phase of development
  348. - **Repetitive requests** - Asking for the same workflow every single session
  349. ### How This Addresses It
  350. **1. One agent, one job, done extremely well**
  351. Same principle as good code: single responsibility. The deslop-agent cleans. The learn-agent researches. The agnix-agent lints agent configs. No agent tries to do everything. Specialized agents, each with narrow scope and clear success criteria.
  352. **2. Pipeline with gates, not a monolith**
  353. Same principle as DevOps. Each step must pass before the next begins. Can't push before review. Can't merge before CI passes. Hooks enforce this - agents literally cannot skip phases.
  354. **3. Tools do tool work, agents do agent work**
  355. If static analysis, regex, or a shell command can do it, don't ask an LLM. Pattern detection uses pre-indexed regex. File discovery uses glob. Platform detection uses file existence checks. The LLM only handles what requires judgment.
  356. **4. Agents don't need to know how tools work**
  357. The slop detector returns findings with certainty levels. The agent doesn't need to understand the three-phase pipeline, the regex patterns, or the analyzer heuristics. Good tool design means the consumer doesn't need implementation details.
  358. **5. Build tools where tools don't exist**
  359. Many tasks lack existing tools. Multi-pass analyzers for slop detection. A rule engine for agent configs. The result: agents receive structured data, not raw problems to figure out.
  360. **6. Research-backed prompt engineering**
  361. Documented techniques that measurably improve results:
  362. - **Progressive disclosure** - Agents see only what's needed for the current step
  363. - **Structured output** - JSON between delimiters, XML tags for sections
  364. - **Explicit constraints** - What agents MUST NOT do matters as much as what they do
  365. - **Few-shot examples** - Where patterns aren't obvious
  366. - **Tool calling over generation** - Let the model use tools rather than generate tool-like output
  367. **7. Validate plan and results, not every step**
  368. Approve the plan. See the results. The middle is automated. /foreman picks, briefs, reads the review and lands without a prompt at each step.
  369. **8. Right model for the task**
  370. Match model capability to task complexity:
  371. - **opus** - Planning, implementation, review orchestration
  372. - **sonnet** - Pattern matching, validation, discovery
  373. - **haiku** - Git operations, file moves, CI polling
  374. Quality compounds. Poor exploration → poor plan → poor implementation → review cycles. Early phases deserve the best model.
  375. **9. Persistent state survives sessions**
  376. State files track the workflow phase. Sessions can die and resume. Multiple sessions run in parallel in separate worktrees.
  377. **10. Delegate everything automatable**
  378. Agents don't just write code. They:
  379. - Clean their own output (deslop-agent)
  380. - Work the issue queue from pick to land ([/foreman](#foreman))
  381. - Lint their own configs ([/agnix](#agnix))
  382. - Research what they need to know ([/learn](#learn))
  383. If it can be specified, it can be delegated.
  384. **11. Orchestrator stays high-level**
  385. The main workflow orchestrator doesn't read files, search code, or write implementations. It launches specialized agents and receives their outputs. Keeps the orchestrator's context window available for coordination rather than filled with file contents.
  386. **12. Composable, not monolithic**
  387. Every command works standalone. [`/deslop`](#deslop) cleans code without needing [`/foreman`](#foreman). [`/agnix`](#agnix) lints configs without needing the full workflow. Pieces compose together, but each piece is useful on its own.
  388. ### What This Gets You
  389. - **Run multiple sessions** - Different tasks in different worktrees, no interference
  390. - **Fast iteration** - Approve plan, check results, repeat
  391. - **Stay in the interesting parts** - Policy decisions, architecture choices, edge cases
  392. - **Minimal review burden** - Most issues caught and fixed before you see the output
  393. - **No repetitive requests** - The workflow you want, without asking each time
  394. - **Scale horizontally** - More sessions, more tasks, same oversight level
  395. </details>
  396. ---
  397. ## Installation
  398. ### Claude Code (Recommended way)
  399. ```bash
  400. /plugin marketplace add agent-sh/agentsys
  401. /plugin install foreman@agentsys
  402. ```
  403. ### All Platforms (npm)
  404. ```bash
  405. npm install -g agentsys && agentsys
  406. ```
  407. Interactive installer for Claude Code, OpenCode, Codex CLI, Cursor, and Kiro.
  408. ```bash
  409. # Non-interactive install
  410. agentsys --tool claude # Single tool
  411. agentsys --tool cursor # Cursor (~/.cursor: skills, commands, rules)
  412. agentsys --tool kiro # Kiro (~/.kiro: prompts, skills, agents)
  413. agentsys --tools "claude,opencode" # Multiple tools
  414. agentsys --development # Dev mode (bypasses marketplace)
  415. agentsys install deslop # One plugin, or plugin:component for one agent, skill or command
  416. ```
  417. [Full installation guide →](./docs/INSTALLATION.md)
  418. ---
  419. ## Requirements
  420. **Required:**
  421. - Git
  422. - Node.js 18+
  423. **For GitHub workflows:**
  424. - GitHub CLI (`gh`) authenticated
  425. **For GitLab workflows:**
  426. - GitLab CLI (`glab`) authenticated
  427. **For repo-intel data (read by deslop):**
  428. - [agent-analyzer](https://github.com/agent-sh/agent-analyzer), downloaded automatically on first use
  429. **For /foreman:**
  430. - Node.js 22+, `gh` authenticated, and `waitfor` on `PATH`
  431. **For /agnix:**
  432. - [agnix CLI](https://github.com/agent-sh/agnix) installed (`npm install -g agnix`, `cargo install agnix-cli`, or `brew tap agent-sh/agnix && brew install agnix`)
  433. **Local diagnostics (optional):**
  434. ```bash
  435. npm run detect # Platform detection (CI, deploy, project type)
  436. npm run verify # Tool availability + versions
  437. ```
  438. ---
  439. ## Research & Testing
  440. The system is built on research, not guesswork.
  441. **Knowledge base** (`agent-docs/`): 8,000 lines of curated documentation from Anthropic, OpenAI, Google, and Microsoft covering:
  442. - Agent architecture and design patterns
  443. - Prompt engineering techniques
  444. - Function calling and tool use
  445. - Context efficiency and token optimization
  446. - Multi-agent systems and orchestration
  447. - Instruction following reliability
  448. **Testing:**
  449. - Jest suite runs on Linux and Windows in CI
  450. - Cross-platform validation (Claude Code, OpenCode, Codex CLI, Cursor, Kiro)
  451. **Methodology:**
  452. - Certainty levels derived from pattern analysis across repositories
  453. ---
  454. ## Documentation
  455. | Topic | Link |
  456. |-------|------|
  457. | Installation | [docs/INSTALLATION.md](./docs/INSTALLATION.md) |
  458. | Cross-Platform Setup | [docs/CROSS_PLATFORM.md](./docs/CROSS_PLATFORM.md) |
  459. | Usage Examples | [docs/USAGE.md](./docs/USAGE.md) |
  460. | Architecture | [docs/ARCHITECTURE.md](./docs/ARCHITECTURE.md) |
  461. ### Workflow Deep-Dives
  462. | Workflow | Link |
  463. |----------|------|
  464. | /foreman Flow | [docs/workflows/FOREMAN.md](./docs/workflows/FOREMAN.md) |
  465. ### Reference
  466. | Topic | Link |
  467. |-------|------|
  468. | Slop Patterns | [docs/reference/SLOP-PATTERNS.md](./docs/reference/SLOP-PATTERNS.md) |
  469. | Agent Reference | [docs/reference/AGENTS.md](./docs/reference/AGENTS.md) |
  470. ---
  471. ## Support
  472. - **Issues:** [github.com/agent-sh/agentsys/issues](https://github.com/agent-sh/agentsys/issues)
  473. - **Discussions:** [github.com/agent-sh/agentsys/discussions](https://github.com/agent-sh/agentsys/discussions)
  474. ---
  475. MIT License | Made by [Avi Fenesh](https://github.com/avifenesh)
  476. ## Contributor instructions
  477. Repository instructions live in AGENTS.md. Claude Code v2.1.281 or later supports
  478. native loading across supported providers. If an ancestor instruction file masks
  479. AGENTS.md, select `claude-md-and-agents-md` in user-level Project instructions.
  480. The platform helpers retain support for analyzing older instruction layouts;
  481. they do not control the native CLI loader.
  482. See [Claude Code instruction loading](https://code.claude.com/docs/en/memory#agents-md).

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.