What is the LISA GitHub Action?

LISA stands for Leak, Injection & Simplicity Auditor. It is a GitHub Action that reviews every pull request with TypeSafe's Jev model and fails the check when it finds a problem, asking four yes/no questions about each part of the diff.

Which problems does it catch?

Secrets (API keys, tokens, passwords, private keys, connection strings), security vulnerabilities (injection, XSS, weakened access control, unsafe eval, SSRF and more), unneeded complexity, prompt injection attempts, and duplicated logic across look-alike files.

How do I add it to a repository?

Get a key at console.typesafe.ai/keys, add it as an Actions secret named TYPESAFE_API_KEY, add the Lisa workflow in .github/workflows with the action pinned to a full commit SHA, and make Lisa a required status check in branch protection.

Why does Lisa fail on fork pull requests?

GitHub does not give pull_request workflows from forks access to secrets, so Lisa fails with Missing api-key. The README suggests on: pull_request_target instead, which is safe only because Lisa never checks out or runs PR code — never add a checkout step to that job.

How does the prompt injection check work?

It looks for text that tries to manipulate an AI system: instructions aimed at AI reviewers or agents, and hidden text. The PR title and description are checked too, not just the diff, and Lisa never checks out or runs PR code while doing it.

Does it need any dependencies at run time?

No. Lisa runs on the Python standard library alone and installs nothing at run time. It reads the diff through the GitHub API, which is also why the README says to keep it in its own job with no other steps.

Which files get which checks?

Every file is checked for secrets and prompt injection. Prose such as Markdown and text gets only those two. Config, data files and generated code also get the security check, and only hand-written code gets the quality checks, complexity and duplication.

Jev directory

CommunityJev Integrations & Adapters

LISA — Leak, Injection & Simplicity Auditor

A GitHub Action that reviews every pull request with TypeSafe Jev and fails the check when it finds a problem. It asks four yes/no questions about each part of the diff and compares files that look duplicated, catching secrets, vulnerabilities, needless complexity and prompt injection.

See highlights Open github.com

Stars: 1
Language: Python
License: MIT
Reading time: 3 min
Published by: Community
Author: turenlabs
Last verified: Oct 2026

turenlabs/lisa

Organization repository on GitHub

View on GitHub

LISA: Leak, Injection & Simplicity Auditor. GitHub Action that uses TypeSafe Jev (a system one model) to flag secrets, security vulnerabilities, and unneeded complexity in pull requests

Forks: 0
Last pushed: Updated Oct 2026

Repo stats from the GitHub API, cached Oct 2026.

Tagscommunity catalog-import yibie github-action security sast

Works forJev for Verification Jev for Moderation Jev for Decision gating

Used inJev for Software & DevTools Jev for Security & Privacy

Languages: Python
Capabilities: Open sourceRequires API key

Highlights

Five checks: secret, security, complexity, prompt_injection and duplication, the last asked about pairs of look-alike files rather than chunks.
Runs on the Python standard library alone and installs nothing at run time; no checkout step — it reads the diff through the GitHub API.
Every file is checked for secrets and prompt injection; prose gets only those two, and only hand-written code gets complexity and duplication.
Findings comment on the exact line and one summary comment stays up to date; maintainers can dismiss findings they have reviewed.
Repositories can add their own questions about a diff chunk or the whole PR, turn checks off, and skip paths in .lisa.toml.

Watch out

Copy reviewed against the repo README on 2026-10-15; the README's inputs table is truncated in the digest and no accuracy benchmarks are published. The README warns that fork pull requests get no secrets, so Lisa fails with Missing api-key unless you use pull_request_target — and warns never to add a checkout step to that job.

Questions about this entry

Short answers drawn from the linked project or page.

Reactions & coverage

Posts, threads, and videos about this entry from around the web.

YouTube: Kubernetes admission control with a decision model, fail-closed.

Hacker News: SafeDep Gryph plus Jev checks for compromised coding-agent behavior.

2 pointsDetecting Compromised AI Coding Agents with Jev and GryphRead the thread on Hacker News

X: Jev and Laya on security queries

X: Testing Jev for prompt injection

From the Jev guides

Original write-ups on what this entry is for.

Decision gates in agent frameworks: what each integration ships

Oct 16, 2026

How Pydantic AI, Mastra, TanStack AI, DSPy, Spring AI, and LiteLLM expose Jev: where the typed decision sits in the loop, what each integration returns, and which big frameworks are only adjacent.

Jev Integrations & Adapters#agents#frameworks#integrations

Read article

From the Jev AI model community

Posts from builders shipping with Jev right now.

Follow @typesafeai