Paladin
Security review for pull requests, repositories, and mitigation planning.
Paladin is a local skill plugin for AI-native engineering teams. It reviews code where the repo context is available, checks current advisory sources when needed, and turns real security findings into clear next steps.
Use it before merging a risky PR, during a weekly security sweep, or when a CVE, dependency alert, scanner result, or review note needs a practical fix plan.
Paladin is defensive by design. It does not promise full coverage, replace a security team, or claim to find unknown zero-days. It reviews known high-risk vulnerability classes, known exploited weaknesses, and regressions against a security baseline.
Why Paladin Runs In Your Repo
Security review depends on context. A generic warning is not enough.
Paladin checks the code paths around a change so it can answer the question that matters:
Is this actually exploitable here, and what should we do next?
Running inside the repo gives Paladin access to:
- changed files, staged files, branches, and PR diffs
- existing auth, permissions, tenant boundaries, validation, escaping, and database patterns
- package manifests, lockfiles, deployment files, CI config, and tests
- local instructions such as
AGENTS.md, CLAUDE.md, Cursor rules, and PALADIN.md
- current Git state and the open PR for the branch
That context helps Paladin avoid noisy reports. If existing authorization, validation, escaping, parameterization, or package versions already block the issue, Paladin should say that instead of inventing work.
What Paladin Does
Paladin gives teams three practical security workflows:
| Workflow | What you get |
|---|
| PR review | A focused review of changed code for medium, high, and critical security findings |
| Repo audit | A prioritized local backlog for risky areas in the codebase |
| Mitigation planning | A concrete fix plan for a known finding, CVE, CWE, dependency alert, or scanner result |
It checks for issues around:
- authentication and session handling
- authorization, ownership checks, and tenant isolation
- user input, file handling, path traversal, SSRF, XSS, request forgery, and injection
- secrets, insecure logging, and leaky error handling
- dependencies, build scripts, supply chain risk, and known exploited vulnerabilities
- infrastructure, deployment config, CORS, headers, and exposed admin paths
Every real finding should include evidence, impact, mitigation, a required regression test, severity, and whether it should block the PR.
Start Here
1. Install Paladin
Run this from the repo where you want Paladin available:
npx skills add onehorizonai/paladin
Registry URL after publication: skills.sh/onehorizonai/paladin
For Claude Code:
/plugin marketplace add onehorizonai/paladin
/plugin install paladin
/reload-plugins
Manual project install:
git clone --depth 1 https://github.com/onehorizonai/paladin .claude/skills/paladin
Manual user install:
git clone --depth 1 https://github.com/onehorizonai/paladin ~/.claude/skills/paladin
Restart your agent tool after installing.
2. Configure The Repo
Run:
/paladin-setup
paladin-setup creates or updates PALADIN.md in the repo root. It asks what Paladin should do after it finds actionable security work:
- create an item automatically
- ask before creating an item
- only report in chat
It also asks where the follow-up should go. One Horizon is the default, but not the only option. You can route findings to Linear, Jira, email, or a custom workflow.
Setup uses repo-root references/security-sources.md as the default advisory source list and creates it from Paladin's bundled source list when the target repo is missing it.
3. Run Paladin
Start with the dispatcher when you want Paladin to choose the right mode:
/paladin
Or call a workflow directly:
/paladin-pr-review
/paladin-repo-audit
/paladin-mitigate
Plain English works too:
Use Paladin to review my uncommitted changes.
Use Paladin to review this PR for security risks.
Use Paladin to audit this repository for security risks.
Use Paladin to turn this CVE into a mitigation plan.
Common Ways To Use Paladin
| Moment | Run |
|---|
| You start work with changed files | /paladin |
| You are about to open or merge a PR | /paladin-pr-review |
| A PR touches auth, permissions, input handling, dependencies, infra, logs, or file handling | /paladin-pr-review |
| You want a weekly security sweep | /paladin-repo-audit |
| A scanner, dependency alert, CVE, CWE, or review note appears | /paladin-mitigate |
Useful cloud-agent prompts:
Before opening a PR, run Paladin on the diff. Include evidence, impact, mitigation, required test, severity, and blocking status.
