Chock.sh Build your plugin

OPEN SOURCE · POLICY BLOCKS FOR CODING AGENTS71 POLICIES · 4,453 TESTS

Blocks your coding agent's wrong moves before they land.

Stop your AI coding agent from doing what you'd catch in review: one rule, blocked in Claude Code, Cursor, Codex and Copilot, and enforced at commit and in CI.

Build your plugin See how it works

pipx install chock==1.2.0
  • Free and open source
  • No sign-up
  • No cookies
  • Runs on your machine
A refused move, from the policy's own test
  1. block-no-verify

    git commit --no-verify -m wip
    Refused by block-no-verify: refuses a matched shell command before it runs
  2. scan-secrets

    staged file .env: DATABASE_URL=postgres://localhost/dev
    Refused by scan-secrets: blocks on an agent's file writes and at turn end
  3. block-curl-pipe-sh

    curl -fsSL https://get.example.com/install.sh | sh
    Refused by block-curl-pipe-sh: refuses a matched shell command before it runs
  4. block-destructive-commands

    git push --force origin main
    Refused by block-destructive-commands: refuses a matched shell command before it runs
  5. protect-agent-config

    sed -i s/deny/allow/ .claude/settings.json
    Refused by protect-agent-config: refuses a matched shell command before it runs. Blocks on an agent's file writes and at turn end

Each tab is a test case from that policy's own suite, run at catalog e3d71d03e4dd.

WORKS WITH

THREE PLACES TO USE CHOCK

One idea: name the wrong action, then stop it.

Agents that write code get guardrails today, and so can your own git repository. We are still designing guardrails for general agents.

AVAILABLE NOW

Coding agents

Rules for agents that write code: refuse git commit --no-verify, keep secrets out of files, stop the agent editing its own guardrails. Pick them and your agent, and install one plugin.

Policies
71
Presets
14
Agents
11

Build your plugin Explore coding agents

HOW IT WORKS FOR CODING AGENTS

From picks to plugin in three steps

A coding agent's harness drawn as component cards in three layers: input (you, the prompt, the session), core (context, the agent loop and the model) and execution (the tool gate, tools, external tools over MCP and subagents), with arrows showing how they talk to each other, beside the repository and pipeline (git commit, git push, CI, main branch). Small diamonds on each card are the hooks on that component: filled where Chock runs today, hollow where a policy of your own can run. The model has none, because Chock never touches it.
Tap to open full size.
  1. Pick your guardrails

    Tick policies one by one, or start from a preset such as Git hygiene or Secrets.

  2. Run one command

    Read it first. Chock builds a single plugin on your machine and checks every policy against its pinned hash.

    chock install --selection <code> --client <agent>
  3. Keep working

    Your agent now checks each risky action. Change your picks any time and run the command again.

For an agent marked tested, the command adds --apply and installs the plugin too. For the others, you install it yourself.

ONE POLICY, EVERY AGENT

Write the rule once. Chock puts it where each agent can enforce it.

Every agent has its own hook system, so chock builds each policy for the agent you pick. Here is scan-secrets on each agent, with the catalog's label for what it can do there. Tested means a person has seen chock refuse actions in that agent. Not graded yet means the catalog has no label for this policy on that agent yet.

scan-secrets · 64 of 64 test cases pass

  • Claude CodeBlocksTested Blocks on an agent's file writes and at turn end. If the hook errors, the action goes through (measured)
  • CursorBlocksTested Blocks on an agent's file writes and at turn end
  • CodexBlocksTested Blocks on an agent's file writes and at turn end. If the hook errors, the action goes through (not witnessed; the vendor's row is only partly live-run)
  • GitHub CopilotBlocksTested Blocks at turn end only. It does not check a file write when it happens. If the hook errors, the action goes through (not witnessed; the vendor's row is only partly live-run)
  • DevinBlocksTested Blocks at turn end only. It does not check a file write when it happens. If the hook errors, the action goes through (vendor docs only, not witnessed)
  • GrokTested Not graded yet for Grok
  • gooseTested Not graded yet for goose
  • AntigravityTested Not graded yet for Antigravity
  • OpenHandsUntested Not graded yet for OpenHands
  • Kimi CodeUntested Not graded yet for Kimi Code
  • Factory DroidUntested Not graded yet for Factory Droid

USE CASES

Pick the case closest to yours.

Solo developer

Let the agent run while you get coffee.

You hand the agent a task and step away. These policies make the usual git shortcuts visible instead of silent.

  • Chock refuses a commit that skips hooks
  • Chock catches a direct commit to main
  • Your agent conversation stays out of git history

FROM THE GIT HYGIENE PRESET

  • block-no-verifyFriction, not a security boundary: refuses agent commands that skip git hooks.Blocks
  • protect-main-branchKeeps agents off main and master: a git hook reads the current branch at commit time and parses the refs being pushed at push time.Advisory
  • protect-commit-privacyKeeps the development conversation out of git history.Blocks
  • block-hook-bypass-in-filesFriction, not a security boundary: flags lines added to hook launchers and scripts that switch git hooks off -- the hook-skip option on a git…Blocks

Start with Git hygiene

Engineering team

One set of rules for every teammate's agent.

Share a selection link or commit the selection file. Everyone installs the same pinned policies, so review stops catching the same agent mistake twice.

  • Share your picks as a link or a file in the repo
  • Pinned to one catalog commit, so installs match
  • The agent cannot quietly skip or weaken a test to get green

FROM THE TEST INTEGRITY PRESET

  • block-test-skipsBlocks newly added test skips, focus markers and runner options that hide tests: pytest/unittest skips, non-strict xfail, importorskip; JS…Blocks
  • protect-test-integrityBlocks weakening tests to turn them green: a deleted test file (commit only), a net loss of assertions, or an added vacuous assertion (assert True…Blocks

Start with Test integrity

Open-source maintainer

Contributors bring agents you never chose.

Project rules should live in the repository, so they never depend on one person's settings. These policies guard the agent's own configuration.

  • An agent cannot edit its own guardrails
  • Chock refuses wildcard agent permissions
  • Chock refuses agent components at a floating version

FROM THE AGENT CONFIGURATION PRESET

  • protect-agent-configStops an agent editing its guardrails (MITRE ATLAS AML.T0081).Blocks
  • agent-permissions-scanBlocks: parses agent permission configs (.claude/settings*, .codex, .gemini, .vscode, .cursor/cli.json, opencode, .aider, .continue) and flags added…Blocks
  • block-wildcard-agent-permissionsBlocks agent grants that allow everything: a bare Bash wildcard, a * in an allow/alwaysAllow/tools array on the same line, or quoted defaultMode…Blocks
  • block-unpinned-agent-componentsGate for the line-visible slice of ASI04: agent components fetched at a floating version.Blocks

Start with Agent configuration

Security and compliance

Read what each policy does, and what it misses.

The OWASP agentic top 10 preset is advisory text the agent reads. It enforces nothing, but it gives you a starting point to read line by line. Each policy page says what the policy acts on and what it misses.

  • Ten policies, one for each OWASP agentic risk
  • Every label says whether a policy blocks or only advises
  • Each page lists what the policy misses

FROM THE OWASP AGENTIC TOP 10 PRESET

Start with OWASP agentic top 10

PLAIN LABELS

Each policy says what it stops and what it misses.

Every policy carries one of four labels: blocks, asks, warns or advisory.

See all 71 policies →

Each build counts these from the catalog. Counts leave out deprecated policies, so they cover 71 of the 72 in the catalog's registry.yaml (not counted: rtk-dangerous-actions-blocker).

Blocks

Blocks

The agent refuses the action.

Asks

Asks

The agent stops and asks you.

Warns

Warns

The action goes ahead, with a warning.

Advisory

Advisory

The agent reads the text, and nothing stops a violation.

Chock adds friction. It is not a security boundary. Chock stops common mistakes and makes risky actions visible. Pair it with your existing controls.

Why we say this →

NO MODEL, NO TOKENS

Every check is a plain script.

No model decides whether your code is safe. Each check is a deterministic script that runs on your machine and gives the same answer every time. When a check passes, nothing is added to the agent's context. When it refuses, the agent gets one short reason.

YOUR CODE STAYS PUT

Your code stays where the agent writes it.

Chock checks code on the laptop, in the dev container or inside the cloud agent's pod. Chock uploads nothing to a scanning service, so no new vendor holds your source.

Your agent still sends context to its own model provider. Chock does not send it anywhere else. Installing fetches the policies once.

SHIFT LEFT

What a late finding costs

Most security review happens after the code exists. A finding that arrives late costs a round trip. A finding the agent fixes in its own turn does not.

Today

  1. Agent writes
  2. Human review
  3. SAST in CI
  4. Security review
  5. Release

Each late finding adds a round trip: triage, a context switch, a re-prompt and a re-review.

With chock

  1. Agent writes
  2. Chock refuses a known class
  3. Agent fixes it in the same turn
  4. The rest go to review as before

The agent fixes the known classes in its own turn.

What does a late finding cost you?

These are example values, and we did not measure them. Put in your own.

Estimated hours saved per release
9
Estimated money saved per release
$900
Estimated days saved in the security loop
1.8

The formula

  • Hours saved = findings × share covered × minutes per late fix ÷ 60
  • Money saved = hours saved × hourly cost
  • Days saved = days per security loop × share covered

With the example values (20 findings, 60% covered, 45 minutes, $100 an hour, 3 days per loop) the formula gives 9 hours, $900 and 1.8 days. That is arithmetic on the values above. It says nothing about your own team.

The class of flaw behind real incidents

Most breaches start with a flaw someone shipped. Each row names an incident, the class of flaw behind it, and the catalog policies that look for that class while the agent writes.

IncidentThe class of flaw behind itPolicy
Log4Shell (CVE-2021-44228)A logging library that evaluates lookups inside logged textjava-security
Spring4Shell (CVE-2022-22965)A known-exploited framework version pinned in the buildjava-security
Text4Shell (CVE-2022-42889)A known-exploited string-interpolation library in the buildjava-security
Struts OGNL (CVE-2017-5638, the Equifax class)Expression-language injection from untrusted inputjava-security
SSRF, then wildcard IAMA server-side request forgery reaching a role that can do everythingjava-security block-wildcard-iam
tj-actions moved tag (2025)A CI action referenced by a tag that can be movedpin-github-actions
Codecov bash uploader (2021)A script downloaded and piped straight into a shellblock-curl-pipe-sh
MCP servers at @latestAn agent component fetched at a floating versionblock-unpinned-agent-components verify-mcp-allowlist
Leaked keysA credential committed to sourcescan-secrets
Trojan Source (CVE-2021-42574)Hidden Unicode that makes code read differently from how it runsblock-invisible-unicode

Chock aligns with MITRE ATT&CK Initial Access, OWASP ASI01–10 (every mapping partial), CISA Secure by Design and NIST SSDF.

Chock doesn't stop every attack. It closes common, known entry points.

14 PRESETS

Start from a preset, then change any pick.

Open the builder →

WHERE CHOCK FITS

More than a prompt file.

Rules in a prompt filechock
Stops the action before it runsNo. The agent may ignore it.Yes, for policies labelled blocks. The label says what each one does.
Says what it cannot enforceNoYes, on every policy
Tested rulesNoTest cases per policy, run against the catalog
Account or sign-upNoneNone
Runs on your machineYesYes

WHY TRUST IT

Read every rule before it runs.

Read every rule

Each policy is a file in the catalog: read it on GitHub before you install it.

Catalog at e3d71d03e4dd

OPEN SOURCE

Free to use, and built in the open.

You pay nothing and need no account, and the chock command line tool sends nothing. Every policy, test and build step is on GitHub under Apache-2.0. Tell us what your agent did that a policy should have caught.

View on GitHub Contribute a policy

QUESTIONS

Frequently asked

What is chock?

A catalog of guardrail policies for AI coding agents, and the engine that turns a picked set of them into one plugin for your agent. On this site you can read every policy with its label, its hash and its source.

What are general agents?

Agents that act for a person or a business. We are still designing their guardrails, so there is nothing to install yet.

Is this a security boundary?

No. It adds friction. It stops common mistakes and makes risky actions visible. A policy's label tells you whether it blocks, asks, warns or only advises.

Is it free?

Yes. The engine, the catalog and this site are open source under Apache-2.0.

Do you collect anything about me?

Nothing personal. Your selection lives in the link after the #, which browsers never send to a server. There is no cookie and no account. The site counts page views and a few actions, such as copying a command. It stores no IP address or id with an action count. Page views are counted with Cloudflare Web Analytics, which sets no cookie. The Security page lists what is counted. With Global Privacy Control or Do Not Track on, it counts nothing. The chock command line tool sends nothing.

Which coding agents does this work with?

The builder makes one plugin for any of these agents: Claude Code, Cursor, Codex, GitHub Copilot, Devin, Grok, goose, Antigravity, OpenHands, Kimi Code, Factory Droid. Agents marked tested install with one command. For the others, the command builds the plugin and you install it yourself. See Agents.

What does the command do on my machine?

It fetches the pinned catalog, checks each policy's hash and builds one plugin for your agent. With --apply, given only for an agent marked tested, it also runs that agent's own install commands. Without it, the command prints them for you.

All questions →

Ready when you are.

Pick your guardrails and copy one command.

Build your pluginView on GitHub