Skip to content
Back to search

grimoire

An agentic auditing stack

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/joranhonig--grimoire/badge.svg)](/listings/joranhonig--grimoire)

HTML

<a href="/listings/joranhonig--grimoire"><img src="/api/listings/joranhonig--grimoire/badge.svg" alt="findsafeskills safety rating" /></a>

Install

i

Install instructions from the README

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

Grimore is set up as a claude plugin that's easy to set up.

At first you should just use the summon command and get going, grimoire will automatically hop in once you ask claude to write a finding, proof of concept, or a similar task.

Once you're ready to dive deep we would suggest having a look at the scribe and cartography skills.

* **scribe** - automatically analyzes your findings and build detection modules for them (which are automatically ran in your next audit) * **cartography** - super useful for large codebases. The cartography skill has claude memorize where to find the context related to different flows/ features. That makes it super easy to load different contexts.

Installation

git clone https://github.com/JoranHonig/grimoire.git
claude --plugin-dir /path/to/grimoire

With Nix, the flake's home-manager module installs Grimoire for the coding agents you pick (`claude-code`, [omp](https://github.com/can1357/oh-my-pi)):

imports = [ grimoire.homeManagerModules.default ];
programs.grimoire = { enable = true; targets = [ "claude-code" "omp" ]; };

For Claude Code this registers the `grimoire` marketplace and enables `grimoire@grimoire` through home-manager's `programs.claude-code` (which must be enabled), so don't also pass `--plugin-dir`. Set `programs.claude-code.mutableSettings = true` to keep managing other plugins with `/plugin`. For omp it links skills and agents only; install `omp-hooks-plus` for hooks.

Without Nix, the repo is also a marketplace: `claude plugin marketplace add JoranHonig/grimoire && claude plugin install grimoire@grimoire`.

Skills and agents auto-discover via the plugin manifest. Some features require API keys:

| Service | Key | Purpose | |---------|-----|---------| | [Solodit](https://solodit.xyz) | `SOLODIT_API_KEY` | Audit findings search via claudit | | [Context7](https://context7.com/dashboard) | `CONTEXT7_API_KEY` | Library documentation lookups |

Set these in your Claude Code settings (`~/.claude/settings.json`):

{
  "env": {
    "SOLODIT_API_KEY": "your-key-here",
    "CONTEXT7_API_KEY": "your-key-here"
  }
}

Both are optional — the librarian will fall back to web search if they are not set. You can also export them as regular shell environment variables (e.g. in `~/.zshrc`) instead of using `settings.json`.

Summon

Start any engagement with `summon`:

You:      "Summon grimoire on this codebase"
Grimoire: → Analyzes project structure, architecture, integrations
          → Identifies crown jewels and attack surface
          → Writes GRIMOIRE.md contextual map
          → Spawns detection checks across the codebase
          → Surfaces initial findings for triage

Then work naturally — Grimoire's skills trigger from context:

"Map the authentication flow"              → cartography
"Write a PoC for the reentrancy I found"   → write-poc
"Document this as a finding"               → finding-draft
"Review my findings before submission"     → finding-review
"Check for duplicates"                     → finding-dedup

Learn more about grimoire by reading [docs](grimoire/concepts/what%20is%20grimoire.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 · 13 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": "grimoire",
  3. "owner": {
  4. "name": "Joran Honig"
  5. },
  6. "plugins": [
  7. {
  8. "name": "grimoire",
  9. "source": "./",
  10. "description": "Security research toolkit for discovering and remediating vulnerabilities"
  11. }
  12. ]
  13. }
README.md · README · 137 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="grimoire_logo.png" alt="Grimoire" width="300" />
  3. </p>
  4. <h1 align="center">Grimoire</h1>
  5. <p align="center">
  6. <strong>A security research toolkit that learns.</strong><br/>
  7. Grimoire takes the raw agent experience and tunes it for security research. Clean, readable and reproducible PoCs,
  8. automatic static analysis module distillation, and more.
  9. </p>
  10. <p align="center">
  11. <a href="#installation"><img src="https://img.shields.io/badge/Claude_Code-Plugin-8B5CF6?style=flat-square" alt="Claude Code Plugin" /></a>
  12. <img src="https://img.shields.io/badge/version-0.1.0-blue?style=flat-square" alt="Version 0.1.0" />
  13. <img src="https://img.shields.io/badge/license-MIT-green?style=flat-square" alt="License MIT" />
  14. </p>
  15. ---
  16. ## Why Grimoire?
  17. There are many audit agents and vulnerability discovery skills.
  18. These are great, but the real power of agents is in amplifying operator skill. Grimoire embraces that philosophy and implements
  19. several skills that make your agent a better co-auditor.
  20. Some skills (such as the cartography) skill come with a small workflow adaption. However, most features such as the *librarian*
  21. are designed to just *work* within whatever workflow you follow.
  22. > The *librarian* is an agent that looks for documentation and references (e.g. previous audit findings, docs, blog posts, etc.) . It is very focussed on providing only information backed up by reference and keeps the main context clear from large mcp descriptions.
  23. >
  24. > The *cartography* skill provides instructions to claude on how it can document a mapping from features / flows to code locations. This allows you to say `hey load context on the authentication flow`, claude will review the file and very quickly load the relevant context.
  25. ### Philosophy
  26. Grimoire is built on a few hard convictions from real-world security research:
  27. - **Leverage over automation.** Grimoire provides skills that amplify operator skill, though some automation is present this is just to give you space for more research.
  28. - **Ergonomics** Agents are already quite useful, Grimoire provides skills (e.g. cartography) that fine tune agent behaviour to be more adjusted to auditing workflows.
  29. - **Not getting in your way** Grimoire takes the base agent experience and makes it more useful for auditing wihout getting in your way.
  30. ### Alpha
  31. Grimoire is still at a very early stage and under continuous development expect there to be major changes.
  32. ## Getting Started
  33. Grimore is set up as a claude plugin that's easy to set up.
  34. At first you should just use the summon command and get going, grimoire will automatically hop in once you ask
  35. claude to write a finding, proof of concept, or a similar task.
  36. Once you're ready to dive deep we would suggest having a look at the scribe and cartography skills.
  37. * **scribe** - automatically analyzes your findings and build detection modules for them (which are automatically ran in your next audit)
  38. * **cartography** - super useful for large codebases. The cartography skill has claude memorize where to find the context related to different flows/ features. That makes it super easy to load different contexts.
  39. ### Installation
  40. ```bash
  41. git clone https://github.com/JoranHonig/grimoire.git
  42. claude --plugin-dir /path/to/grimoire
  43. ```
  44. With Nix, the flake's home-manager module installs Grimoire for the coding agents you pick (`claude-code`, [omp](https://github.com/can1357/oh-my-pi)):
  45. ```nix
  46. imports = [ grimoire.homeManagerModules.default ];
  47. programs.grimoire = { enable = true; targets = [ "claude-code" "omp" ]; };
  48. ```
  49. For Claude Code this registers the `grimoire` marketplace and enables `grimoire@grimoire` through home-manager's `programs.claude-code` (which must be enabled), so don't also pass `--plugin-dir`. Set `programs.claude-code.mutableSettings = true` to keep managing other plugins with `/plugin`. For omp it links skills and agents only; install `omp-hooks-plus` for hooks.
  50. Without Nix, the repo is also a marketplace: `claude plugin marketplace add JoranHonig/grimoire && claude plugin install grimoire@grimoire`.
  51. Skills and agents auto-discover via the plugin manifest. Some features require API keys:
  52. | Service | Key | Purpose |
  53. |---------|-----|---------|
  54. | [Solodit](https://solodit.xyz) | `SOLODIT_API_KEY` | Audit findings search via claudit |
  55. | [Context7](https://context7.com/dashboard) | `CONTEXT7_API_KEY` | Library documentation lookups |
  56. Set these in your Claude Code settings (`~/.claude/settings.json`):
  57. ```json
  58. {
  59. "env": {
  60. "SOLODIT_API_KEY": "your-key-here",
  61. "CONTEXT7_API_KEY": "your-key-here"
  62. }
  63. }
  64. ```
  65. Both are optional — the librarian will fall back to web search if they are not set. You can also export them as regular shell environment variables (e.g. in `~/.zshrc`) instead of using `settings.json`.
  66. ### Summon
  67. Start any engagement with `summon`:
  68. ```
  69. You: "Summon grimoire on this codebase"
  70. Grimoire: → Analyzes project structure, architecture, integrations
  71. → Identifies crown jewels and attack surface
  72. → Writes GRIMOIRE.md contextual map
  73. → Spawns detection checks across the codebase
  74. → Surfaces initial findings for triage
  75. ```
  76. Then work naturally — Grimoire's skills trigger from context:
  77. ```
  78. "Map the authentication flow" → cartography
  79. "Write a PoC for the reentrancy I found" → write-poc
  80. "Document this as a finding" → finding-draft
  81. "Review my findings before submission" → finding-review
  82. "Check for duplicates" → finding-dedup
  83. ```
  84. Learn more about grimoire by reading [docs](grimoire/concepts/what%20is%20grimoire.md)
  85. ## Project Structure
  86. ```
  87. grimoire/ # Human-written specs (read-only, source of truth)
  88. ├── agents/ # Agent specifications
  89. ├── skills/ # Skill specifications
  90. ├── concepts/ # Design philosophy
  91. ├── flows/ # Multi-step workflow specs
  92. └── ideas/ # Roadmap and research notes
  93. skills/ # Implemented skills (where development happens)
  94. agents/ # Implemented agents
  95. ```
  96. Specs and implementation are strictly separated. The `grimoire/` directory is the source of truth — never modified during development. Skills in `skills/` implement those specs.
  97. The `grimoire/` directory is also written as a form of documentation, feel free to browse around if you want to learn more.

Ratings and comments

Sign in to rate this skill.

Sign in

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