Agentic Security plugin
A code security plugin: finds vulnerabilities, challenges its findings, proves exploitability and fixes with a verified patch
- 121 ★ on the repository
- Last commit
Broad access
Why this level
- Exploit confirmation executes reproducer code and fixes are written to project files
- Hooks and deploy commands can block releases and change CI configuration
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add plugins/agentic-securityDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- Exploit confirmation executes reproducer code and fixes are written to project files
- Hooks and deploy commands can block releases and change CI configuration
This is third-party code. Review the repository files before installing.
What it does
Agentic Security pairs deterministic code analysis with security agents. Agents hunt for weaknesses and challenge their own findings, while severity, reachability and exploit confirmation come from mechanical checks, not from the model. An unfinished or partial scan is never reported as clean. In Claude Code the plugin adds slash commands: `secure` suggests the next step, `scan` runs the scanner, `triage` works through findings, `fix` prepares a patch, `supply` checks the supply chain, `image` inspects container images and `compliance` gathers technical evidence against framework controls. A patch is written only after a rescan, lint and tests, and the report names the checks that did not run. The same engine runs from a terminal with `npx @clear-capabilities/agentic-security-scanner`. Reports come out as HTML, JSON, Markdown, SARIF and OSCAL. The repository includes a deliberately vulnerable demo app for a first run.
Who it is for. Development and security teams that want an agent to check their code and images and return evidence rather than a list of guesses.
Good fit when
- Before a release you scan a project and get findings with reachability assessed
- A finding looks doubtful and you want to check whether it reproduces in an isolated run: `triage exploit <id> --prove`
- An agent fixes a vulnerability and, before the patch is written, a rescan, lint and tests must pass
- CI needs a gate that also fails when the scan itself did not finish: `ci --assurance strict`
- You need an SBOM, a container image check or an OSCAL report against framework controls
Not a fit when
- You need a guarantee of no vulnerabilities or a compliance certificate: compliance reports give technical evidence only
- You need to replace a manual pentest or human review: the author calls the tool a complement to them
- You want to resell the scanner or build it into a competing product: the PolyForm Internal Use license forbids it
Example request
Run /agentic-security:secure in this project and show what to do next.Limitations
Node.js 24 or newer is required. The PolyForm Internal Use license permits internal use and forbids resale, and GitHub does not detect it automatically. The author also publishes his own accuracy measure on code the engine was not built from: localized precision 6 of 19, recall 6 of 278, so a clean scan does not prove safety. Some capabilities are labelled Preview, Labs or Roadmap and are less stable. Fixes and checks run on your machine, and exploit confirmation executes reproducer code.
How to disable. Remove the plugin with /plugin in Claude Code and the global package with npm uninstall -g @clear-capabilities/agentic-security-scanner. Delete the .agentic-security directory with reports by hand.
FAQ
Where do I start after installing?
Run `/agentic-security:secure`. It inspects the project and tells you what to run next. The `--tour`, `--help` and `--daily` flags also work.
How can I try it without touching my code?
The repository has a deliberately vulnerable demo app. Run `npx @clear-capabilities/agentic-security-scanner ci examples/demo-app --assurance strict` and read the output.
What does assurance strict mean?
The gate looks not only at what was found but at whether the scan itself finished. If an analyzer is unavailable or data is stale, a `partial` status does not pass as clean. The flag exists only on `ci`.
Does it work outside Claude Code?
Yes. The engine runs from a terminal through `npx @clear-capabilities/agentic-security-scanner`, and the repository documents setup for Codex, Cursor and Gemini CLI.
Which report formats are available?
HTML, JSON, Markdown, SARIF, OSCAL and CSV, set with `--format`, for example `scan . --format html --output report.html`.
Discussion
No questions or reviews yet.
Related
PluginsA Claude Code skill that finds vulnerabilities in one session: threat model, verified findings, CWE and CVSS report.
An AppSec plugin for Claude Code: secret, dependency and code scanners plus skills that judge whether each finding is real
A security scanner for AI-written code: secrets, dependencies, MCP and CI, findings investigated with evidence, and fixes you approve
Prowler plugin for Claude Code
Prowler for Claude Code
A plugin that takes a cloud account toward CIS, PCI-DSS and other frameworks: reads Prowler Cloud reports, fixes failures and proposes a rescan
A pack of about fifty Claude Code subagents for authorized pentesting: planning, recon, exploit research and chaining, post-exploitation, detection and reportin
AgentShield audits the .claude folder for plaintext keys, broad permissions, dangerous hooks and risky MCP servers and grades the setup from A to F
README in short
The platform is built as a ten-stage cycle, from discovery and proof to fix, verification and learning. It explains trust principles, capability maturity, commands by cycle stage, installation and language coverage. Separate sections say what the tool does not do: it issues no certificate and does not replace manual review.