Bonsai-27B-GGUF
lmstudio-community · Kit
A GGUF conversion of the 27B Bonsai model for LM Studio, with a 3.80 GB Q1_0 language file and a 0.93 GB BF16 vision projector. Tested for text, vision and par…
New
Overview
Trail of Bits’ marketplace of more than forty plugins for AI-assisted security work in Claude Code and Codex, from code audit and bug triage to fuzzing, smart contract checks and YARA rules. You install only the plugins you want.
Trail of Bits Skills is an open-source project on GitHub in the Developer category, listed by Mindket. It is free. This is version 2, last updated on . Install it from GitHub as its README says. Get Mindket’s setup notes on your Mindket account, as a zip.
At a glance
- Runs in
- Claude Code: /plugin marketplace add trailofbits/skills, then install plugins from the menu or by name
- Codex: codex plugin marketplace add trailofbits/skills, which works because Codex reads Claude plugin marketplaces, the README says
- ChatGPT workspace: an import of the repository’s .claude-plugin/marketplace.json into your workspace marketplace
- Start with
/plugin marketplace add trailofbits/skills /plugin menu
Platform
Claude Code
Codex
Claude
ChatGPT
About
What it is
Trail of Bits Skills is the security firm’s plugin marketplace for AI coding agents. It holds more than forty plugins, and most carry skills that Claude reads when your request fits. They cover code auditing, verification and testing, smart contracts, malware rules, reverse engineering, mobile checks and developer tooling. Some add slash commands, subagents or hooks. Install only the ones you need.
- You add the Trail of Bits marketplace once
- You install the plugins that fit your work
- Claude reads a skill or you type a plugin command
- It reads your code and writes a report, rules or tests
- You judge each finding before you act on it
Who it suits
Security engineers, auditors and developers who want an AI agent to follow a security firm’s method for reviewing code, triaging bug reports and testing. The README groups the plugins by job, from smart contract security to team management.
| Who | Plugins they reach for | Example |
|---|---|---|
| Code auditors | audit-context-building, differential-review, variant-analysis | An auditor maps an unfamiliar codebase function by function, then reviews a commit for security regressions. |
| Bug triagers | fp-check, vulnerability-triage-brocards, cvss-scoring | A maintainer checks whether a reported bug is real, then gets a CVSS v4.0 vector for it. |
| Smart contract teams | building-secure-contracts, entry-point-analyzer | A team lists a contract’s state-changing entry points and runs the vulnerability scanner for its blockchain. |
| Testing and tooling engineers | static-analysis, semgrep-rule-creator, mutation-testing, property-based-testing | An engineer has Claude write a custom Semgrep rule, or sets up a mutation testing campaign and reads the surviving mutants. |
| Project owners | open-sourcing, modern-python, devcontainer-setup, gh-cli | A maintainer prepares a repository for public release by checking secrets, licensing and CI. |
Where to use it
The README documents three routes for the same marketplace.
- Claude Code:
/plugin marketplace add trailofbits/skills, then install plugins from the menu or by name - Codex:
codex plugin marketplace add trailofbits/skills, which works because Codex reads Claude plugin marketplaces, the README says - ChatGPT workspace: an import of the repository’s
.claude-plugin/marketplace.jsoninto your workspace marketplace
How to set it up
Get it saves a zip of Mindket’s setup notes to your downloads folder, with a page for each route the README gives.
| Need | Detail |
|---|---|
| Claude Code or Codex | For the two command line routes. A ChatGPT workspace uses the import route instead |
| The tools a plugin names | Some plugins call Semgrep, uv, gh, clang, Burp Suite Professional or a Codex, Antigravity or Gemini CLI. Each plugin’s README says which |
| An account or key | None to add the marketplace. second-opinion needs a Codex, Antigravity or Gemini CLI that you sign in to yourself |
- Add the Trail of Bits marketplace. You run this once inside Claude Code.
- Pick the plugins you want. Open the plugin menu, or install one by name, then read the details pane.
- Ask for a job a plugin covers. Claude reads the matching skill, or you type the plugin’s command.
/plugin marketplace add trailofbits/skills /plugin menu
Claude Code’s docs open the plugin panel with /plugin alone, so type that if /plugin menu is refused. To install one plugin by name, use the form Claude Code’s docs give: the plugin, @, then the marketplace name. The plugin READMEs show other forms too, and the kit’s claude-code.md lists them.
/plugin install second-opinion@trailofbits
Codex has its own three lines.
codex plugin marketplace add trailofbits/skills codex plugin list codex plugin add <plugin-name>@trailofbits
How to use it
Run a plugin’s command, or describe the job and let Claude pick the skill. A command starts with the plugin’s name.
| Command | What you give it | What comes back |
|---|---|---|
| /audit-context-building:audit-context | A path, and optionally --focus <module> | Notes on each function written to files, with a dossier, so the conversation stays small |
| /differential-review:diff-review | A pull request address, a commit hash or a diff file, and optionally --baseline <ref> | A markdown report with findings, test coverage gaps, blast radius and regression risks |
| /insecure-defaults:audit | A repository, folder or file, or nothing for the whole repository | Fallback secrets, default credentials, fail-open switches, weak crypto, loose access and debug leaks, each checked by a refuting agent |
| /second-opinion:second-opinion | Plain words naming Codex, Antigravity or Gemini CLI, and what to review | An independent review that names the provider and model, with findings and gaps in coverage |
| /semgrep-rule-creator:semgrep-rule | A bug pattern you want to catch | A custom Semgrep rule, which Claude writes test first |
| /entry-point-analyzer:entry-points | A source folder such as src/core/ | The state-changing entry points of a smart contract codebase, grouped by who may call them |
| /spec-to-code-compliance:spec-compliance | A path such as ./contracts. The workflow finds the specification, or you name it in plain words | A report file that checks each requirement against the code and lists behavior no document mentions |
| /firebase-apk-scanner:scan-apk | An Android APK or a folder of them, for an app you may test | A scan of its Firebase setup for open databases, storage buckets, sign-in problems and exposed functions |
Some plugins have no command and start from a plain request, such as asking fp-check whether a finding is a true positive.
Example
A teammate’s commit changes your login code, and you want it checked for security regressions before it merges.
/differential-review:diff-review 3f9a2c1 --baseline main
Claude sorts the changed files by risk, reads git blame for regressions, counts callers to size the blast radius, and writes a markdown report with findings and test gaps.
For a high risk change the skill hands exploit modeling to its adversarial-modeler agent.
For newcomers and experts
If you are new. A plugin is a bundle that can hold skills, commands and subagents. Start with one plugin that matches a job you have this week, such as differential-review for a code change. Read its report before you trust a finding, since the skill has Claude state its coverage limits and confidence.
For experts. The README says one Claude marketplace file feeds Claude Code, Codex and ChatGPT workspace imports, with no Codex-only copy of the metadata. Some plugins run as workflows. audit-context-building has one subagent analyze each function and write its notes to a file. insecure-defaults runs six parallel sweeps, then refuting agents that try to disprove each finding. Several plugins also ship evals and tests in the repository.
Good to know
Plugins can carry hooks and tools, so read each plugin’s details before you install it.
- Hooks: fp-check checks at stop that its verification phases are complete, gh-cli redirects GitHub fetches to the
ghCLI, and modern-python puts shims on PATH that stop a barepythonorpip - A server: zeroize-audit starts the Serena MCP server through
uvx, fetched from another GitHub repository - Changes to your repositories: github-triage can merge pull requests and close issues, and git-cleanup deletes local branches and worktrees. Both wait for your approval first
- Testing other systems: firebase-apk-scanner sends test requests to the Firebase services an APK points at. Use it only on apps you may test
- Data: second-opinion sends a captured patch to the review tool you pick, and what Claude reads goes to your model provider
- Install lines: the plugin READMEs show several forms, and Claude Code’s docs give the plugin@marketplace form
- Limits: the supply-chain auditor’s report states what it could not measure, and the differential review skill has Claude state its coverage limits
- License: CC BY-SA 4.0, by Trail of Bits. Using the plugins asks nothing of you. If you share them or a changed copy, credit Trail of Bits, link the license and say what you changed. A changed copy you share must carry CC BY-SA 4.0 or a compatible license. The kit holds Mindket’s setup notes and no plugin text
README and files
# Trail of Bits Skills: setup notes This kit holds setup notes written by Mindket. It does not contain any plugin, skill, hook, script or rule file. Those install from Trail of Bits’ repository: https://github.com/trailofbits/skills License: Creative Commons Attribution-ShareAlike 4.0 International (CC BY-SA 4.0), as the README says: “Made by Trail of Bits”. Full text: https://github.com/trailofbits/skills/blob/main/LICENSE. Using the plugins asks nothing of you. If you share them or a changed copy, the license asks you to credit Trail of Bits, link the license and say what you changed. A changed copy you share must carry CC BY-SA 4.0 or a license compatible with it. This kit copies none of the repository’s text. The commands in it are copied from the README and the plugin READMEs, and each one names the file it came from. Trail of Bits Skills is made by Trail of Bits. This listing is by Mindket, which is not made or endorsed by Trail of Bits. The pages these notes came from: the README (https://github.com/trailofbits/skills/blob/main/README.md), the marketplace file (https://github.com/trailofbits/skills/blob/main/.claude-plugin/marketplace.json), AGENTS.md, and the README and SKILL.md files of the plugins named below, under https://github.com/trailofbits/skills/tree/main/plugins. Facts about Claude Code come from its docs, named where they are used. Mindket copied every command and ran none of them. ## What it is A plugin marketplace for AI coding agents with more than forty plugins, grouped in the README by job. Most plugins hold one or more skills, and some add slash commands, subagents, hooks or an MCP server. The repository’s AGENTS.md says one Claude plugin marketplace file is the source for Claude Code, Codex and ChatGPT workspace imports, with no separate Codex copy. ## The plugins, by the README’s groups | Group | Plugins | |---|---| | Smart Contract Security | building-secure-contracts, entry-point-analyzer | | Code Auditing | agentic-actions-auditor, audit-context-building, burpsuite-project-parser, c-review, cvss-scoring, differential-review, dimensional-analysis, fp-check, insecure-defaults, rust-review, semgrep-rule-creator, semgrep-rule-variant-creator, sharp-edges, static-analysis, supply-chain-risk-auditor, testing-handbook-skills, trailmark, variant-analysis, vulnerability-triage-brocards | | Malware Analysis | yara-authoring | | Verification | constant-time-analysis, mutation-testing, patching-bugs, post-patch-validation, property-based-testing, regression-hunt, spec-to-code-compliance, writing-lean-proofs, zeroize-audit | | Reverse Engineering | dwarf-expert | | Mobile Security | firebase-apk-scanner | | Development | code-improver, devcontainer-setup, gh-cli, git-cleanup, goal-prompt, github-triage, let-fate-decide, modern-cpp, modern-python, open-sourcing, review-walkthrough, second-opinion | | Team Management | culture-index | | Tooling | chrome-mcp-troubleshooting | Source: the README’s “Available Plugins” tables, checked against the plugin list in marketplace.json. ## Files in this kit - setup/claude-code.md: the README’s Claude Code route, the install lines the plugin READMEs give, and a local checkout - setup/codex.md: the README’s three Codex commands - setup/chatgpt-workspace.md: what the README says about importing the marketplace into a ChatGPT workspace - setup/plugin-extras.md: the hooks, tools and accounts some plugins need, from their READMEs ## What you need first - Claude Code or Codex, for the two command line routes, or a ChatGPT workspace that can import a marketplace. - Nothing else to add the marketplace. Each plugin may need its own tools: setup/plugin-extras.md. - The README asks for no account or API key to install the marketplace. ## What installs where Claude Code keeps an installed plugin’s files itself. A plugin’s skills run under the plugin’s name, as /plugin-name:skill-name, and its commands run as /plugin-name:command-name, for example /differential-review:diff-review. Codex and ChatGPT workspaces read the same marketplace file. ## Your data The plugins are instructions, commands and helper programs that run on your machine and read the code you point them at. What your agent reads goes to the model provider you chose. A few plugins send code on purpose: second-opinion sends a captured patch to Codex, Antigravity or Gemini CLI, which you set up yourself. Three plugins carry hooks and one starts an MCP server: setup/plugin-extras.md. ## Update and remove The README gives no update or removal step. Claude Code’s plugin page says the Installed tab in `/plugin` lists each plugin with actions to update or uninstall it, and Claude Code’s skills page names `/plugin uninstall <plugin-name>@<marketplace-name>`, here for example second-opinion@trailofbits.
README.md 5 KB setup/chatgpt-workspace.md 1 KB setup/claude-code.md 4 KB setup/codex.md 2 KB setup/plugin-extras.md 4 KB
What’s new
Updated by Mindket to the current version of this kit.
Compare with similar
Listed GitHub projects in the same category that share the most topics with this one.
| Product | Stars | Activity | Language | License | Last change |
|---|---|---|---|---|---|
| Trail of Bits Skills This page | 8,060 | — | — | CC-BY-SA-4.0 | — |
| Anthropic Agent SkillsShares agent-skills | 194,161 | Active | Python | — | |
| PonytailShares agent-skills | 186,952 | Active | JavaScript | MIT | |
| Agent SkillsShares agent-skills | 116,705 | Active | JavaScript | MIT | |
| Taste SkillShares agent-skills | 115,072 | Active | JavaScript | MIT | |
| ImpeccableShares agent-skills | 109,647 | Active | Rust | Apache-2.0 |
Reviews
What people say
Posts and comments that name this project on Hacker News, Bluesky, X, Reddit, Facebook and YouTube, found by a weekly search and on Trendshift. Mindket doesn’t write or check them. Open one to read it where it was posted.
dheerajmpClaude Skills for Cybersecurity3 pointsr1cksecTrail of Bits Claude Code skills for security research, vulnerability detection, and audit workflows github.com/trailofbits/... #infosec #cybersecurity #redteam #pentest #ai5 likeswslhTrail of Bits Skills Marketplace2 pointsiterativeguyskills: Security Workflows for Claude Code and Codex https://osrepos.com/repo/trailofbits-skills?utm_source=bluesky Trail of Bits Skills is a Claude Code plugin marketplace for AI-assisted security analysis, testing, and development. #opensource3 likeswslhTrail of Bits Skills Marketplace2 pointsJames GI hear you. This repo is a good start to understand what’s available github.com/trailofbits/...1 like · 1 replythemaxdavittLet-Fate-Decide2 pointsEveryDev AI3. tstack from @theo, Claude Code and Codex skills for filing, reviewing, babysitting and merging PRs. 4. Brotli Checker from @trailofbits, an agent skill and scanner that finds Brotli decoder copies that may need patching for CVE-2026-111497.1 replysome_furryThe V8 JavaScript Runtime Undermined My Constant-Time JavaScript LibraryYou can write algorithms that leak every bit of your secrets, trivially. The npm "elliptic" package does this. Daniel Bleichenbacher has tested several packages' timing leakage in Rooterberg that are easily exploited: https://github.com/bleichenbacher-daniel/Rooterberg/blob/mai... You can use…Claus Cramon HoumannSo @trailofbits publish some nice research that shows a much higher % fix ratio from Agentic AI/skills than the research by FASCIST company 1Password, they don't dunk on 1pw tho, so let me do it: dunked https://infosec.exchange/@trailofbits/117274765830879355dlahodaAikido Code AuditI run orchestrated agents(context warm up, fork-join, file based communication, set of trailofbits skills, some dba skill, some sourcegraphindex, and few mine, API skill). None of skills are specific to out codebase(even my), some are explicitly for Claude. Finds 90% bugs using Codex and Gemini…Tom Dörrhttps://github.com/trailofbits/skills
More from Mindket
See all from MindketDeepSeek-V4.1-Flash-1017 EXL3 2.9 bpw
LibertAIDAI · Kit
Runs DeepSeek-V4.1-Flash-1017 as a 2.9 bits-per-weight EXL3 quantization, a 196 GiB copy for people serving large models on their own hardware. Perplexity stay…
New
MusicGen Small
facebook · Kit
Generates short music clips from a text description or an audio prompt, in the small checkpoint of facebook's MusicGen. It suits developers and music hobbyists…
New
Similar in Developer
See all in DeveloperLexnet
alpindale · Kit
Tags every character of C source code for syntax highlighting with a tiny neural network of about 984 weights, run through a small C program. For developers an…
New
FLUX.2 Klein LoRA Studio
prithivMLmods · Kit
Try a set of LoRAs for the FLUX.2 Klein image model in a Gradio app that runs in your browser or on your own computer. Made for people who make or test AI imag…
New
anaksor-ocr
Sokheng · Kit
Reads Khmer and English text from documents, receipts and photos of signs, using a detector for text lines and a recogniser for each line. Suits developers who…
New