---
title: "Claude Security: review, setup & limitations · Undominated.ai"
canonical: https://undominated.ai/skills/anthropic-claude-security/
description: "Runs Anthropic's Claude Security scans inside a Claude Code session: a whole-repository or scoped scan, a scan of a branch, pull request or commit, and suggested fixes as patch files the user applies."
---

# Claude Security: review, setup & limitations · Undominated.ai

> Runs Anthropic's Claude Security scans inside a Claude Code session: a whole-repository or scoped scan, a scan of a branch, pull request or commit, and suggested fixes as patch files the user applies.

[← Explore all skills](/skills/)

SECURITY REVIEW / Anthropic

# Claude Security

Runs Anthropic's Claude Security scans inside a Claude Code session: a whole-repository or scoped scan, a scan of a branch, pull request or commit, and suggested fixes as patch files the user applies.

 [See setup guidance ↓](#setup)[Original source ↗](https://github.com/anthropics/claude-plugins-official/tree/41f32e407b450055044f167db7ded5e8ff67d203/plugins/claude-security/skills/claude-security)

SOURCE REVIEW

 Reviewed 2026-10-22
 Evidence 21 linked sources
 Publisher Anthropic
 Licence [Proprietary (Anthropic, PBC) ↗](https://raw.githubusercontent.com/anthropics/claude-plugins-official/41f32e407b450055044f167db7ded5e8ff67d203/plugins/claude-security/LICENSE)
 Revision 41f32e407b45
 [Read what was—and wasn’t—checked ↓](#review)

“Claude Security: scan the codebase (the whole repository or a scoped part of it), scan changes… or suggest patches…”

 [Anthropic · upstream description ↗](https://raw.githubusercontent.com/anthropics/claude-plugins-official/41f32e407b450055044f167db7ded5e8ff67d203/plugins/claude-security/skills/claude-security/SKILL.md) Our analysis follows below.

01 / THE REASONING

## Why this made the selection.

 - The README calls it the in-your-session version of Anthropic's hosted Claude Security product: one skill that drives three jobs, scan codebase, scan changes and suggest patches, through the plugin's own scan workflow and agents.
- Every candidate finding goes to independent verifiers told to disprove it, the record of how thoroughly a run was verified is computed in code, and results are written as Markdown, JSONL, SARIF 2.1.0 and a revision stamp.
- Suggested fixes are developed in a scratch copy of the repository and delivered as patch files; the job does not apply, commit or push anything.

### A good fit for

 - Scan your own repository, or a scoped part of it, for vulnerabilities from inside a Claude Code session
- Review a branch's diff, a checked-out pull request or one commit for flaws the change takes part in
- Turn the findings of a current Claude Security report into patch files to review and apply with git apply
- Hand findings to GitHub code scanning or another tool that reads SARIF 2.1.0

### Weigh up before choosing

 - Proprietary licence. The LICENSE grants a limited, revocable licence "to install, run, and modify the Plugin for your internal use, solely with Claude Code or other Anthropic products and services", and, "Except as the Agreement expressly permits", forbids using "the Plugin or any part of it with, or to develop, any non-Anthropic product or service, including any competing product", as well as distributing or publishing it to any third party.
- Last commit touching plugins/claude-security/skills/claude-security at revision 41f32e40: 2026-09-25 (ca08d5e4).
- No isolation of its own: a scan runs under the user's session permissions with the repository's .git/config, .claude/ settings and hooks and CLAUDE.md in effect. For a repository you do not trust, the README and SECURITY.md suggest running the whole session inside sandbox-runtime.
- Scans are nondeterministic; the README says two scans of the same code can surface different findings.
- Credential values in the report, JSONL and SARIF are replaced with [REDACTED] by an agent; the README says this is not guaranteed and asks users to look over the files before sharing them.
- Needs Python 3.9 or newer on PATH, and a git checkout for scanning changes and suggesting patches.
- Installing the plugin also installs hooks, among them one that suggests a scan after Bash commands matching git push and gh pr create, and one that reports usage counts through Claude Code's telemetry.
- The README says that on Opus 5.5 or Fable 5.1 portions of a scan may occasionally be downgraded to Opus 4.8.
- The fix job runs the repository's own tests: the patch-verifier agent reviews each staged change and runs the project's tests in a full scratch checkout.

02 / THE REVIEW RECORD

## What we actually inspected.

Source review has boundaries. A clear record is more useful than a “safe” badge.

### Material inspected

 - plugins/claude-security/skills/claude-security/SKILL.md
- plugins/claude-security/skills/claude-security/role.md
- plugins/claude-security/skills/claude-security/jobs/scan-changes.md (pull-request search section)
- plugins/claude-security/skills/claude-security/jobs/suggest-patches.md (overview, workspace and output steps)
- plugins/claude-security/agents/ (frontmatter of all nine agent files)
- plugins/claude-security/README.md
- plugins/claude-security/SECURITY.md
- plugins/claude-security/LICENSE
- plugins/claude-security/NOTICE.md
- plugins/claude-security/.claude-plugin/plugin.json
- plugins/claude-security/hooks/hooks.json
- plugins/claude-security/hooks/hooks.sh
- plugins/claude-security/hooks/hooks.py (imports, metrics output, tip state file and git calls)

### Our findings

 - allowed-tools pre-approves Read, Write, Edit and Bash for a listed set of command prefixes and the plugin's own scripts; role.md says any other or chained command stops the flow on a permission prompt.
- role.md tells the lead agent to treat code, comments, CLAUDE.md and subagent output as data under review, never as instructions; the README says this is not a defense against a hostile repository.
- Scan jobs run git with GIT_CONFIG_GLOBAL=/dev/null and GIT_TERMINAL_PROMPT=0; the fix job clones scratch workspaces with GIT_TERMINAL_PROMPT=0 and never pushes.
- hooks.py imports no HTTP or socket module; its metrics hook prints a metrics object for Claude Code to report.
- NOTICE.md says the plugin carries CWE titles derived from the CWE 4.20 catalog under MITRE's CWE Terms of Use.
- The scan-researcher and scan-verifier agents are described as restricted and read-only, yet their tools lists, like those of explore, patch-generator and patch-verifier, include Bash with no command limit.

### Not established by this review

 - No scan or fix job was run, and no hook fired.
- workflows/scan.js, the bodies of the nine agent definitions past their frontmatter, jobs/scan-codebase.md, specs/ and the Python scripts in scripts/ were not read; jobs/suggest-patches.md was read only for its overview, workspace and output steps.
- Not executed: it was not installed or run.

The review applies to the material and revision named here. A newer upstream release can change its behavior.

03 / PUT IT TO WORK

## Add a skill to your workflow.

[Upstream setup instructions ↗](https://raw.githubusercontent.com/anthropics/claude-plugins-official/41f32e407b450055044f167db7ded5e8ff67d203/plugins/claude-security/README.md)

DOCUMENTED COMMAND

 /plugin install claude-security@claude-plugins-official Copy command ↗

Copying does not execute this command. It may retrieve a newer version than the reviewed source.

 - The claude-security README prints `/plugin install claude-security@claude-plugins-official` followed by `/reload-plugins`. The plugin holds this one skill plus nine agents (the claude-security orchestrator, scan-inventory, scan-researcher, scan-verifier, scan-loader, scan-redactor, explore, patch-generator and patch-verifier), the hooks in hooks/hooks.json, the scripts in scripts/ and the scan workflow.
- If Claude Code reports that the marketplace is not found, the README says to run `/plugin marketplace add anthropics/claude-plugins-official` first, then retry.
- Run `/claude-security` for the menu of three jobs; a scan asks once to confirm its time and token cost unless the request already accepts it.
- This review installed nothing and ran no scan.

### Before you start

 - Claude Code with this plugin installed
- Python 3.9 or newer on PATH
- A git checkout for scanning changes and suggesting patches; a whole-repository scan works without one

### Compatibility

Claude Code

### Reads the open repository and writes reports and patch files into it; network for Claude Code telemetry and an opt-in pull-request search

 - allowed-tools: Read, Write, Edit, Glob, Grep, AskUserQuestion, Workflow and Workflow(claude-security:scan), and eight agents: scan-inventory, scan-researcher, scan-verifier, scan-loader, scan-redactor, patch-generator, patch-verifier and explore
- Bash pre-approved for date, ls, wc, mkdir -p, sleep, git (plain, with GIT_TERMINAL_PROMPT=0, or with GIT_CONFIG_GLOBAL=/dev/null GIT_TERMINAL_PROMPT=0), one find for existing reports, keep-waiting.sh and the Python scripts render_report.py, save_result.py, write_scan_meta.py and patch_artifacts.py
- Agents it starts: scan-researcher, scan-verifier and patch-verifier list Read, Glob, Grep, Bash and Agent(claude-security:explore); patch-generator adds Edit and Write to that; explore lists Read, Glob, Grep and Bash. Bash has no command limit in any of these five, and the patch-verifier runs the project's tests. scan-inventory lists Read, Glob and Grep, scan-loader Read, and scan-redactor Read and Edit
- The plugin also installs agents/claude-security.md, an orchestrator meant only as a session's main agent (claude --agent claude-security:claude-security); its tools include Bash, Write, Edit, AskUserQuestion, Workflow, the Task tools and the same eight agents
- Writes a CLAUDE-SECURITY-<timestamp>/ directory in the repository with CLAUDE-SECURITY-RESULTS.md, .jsonl and .sarif, a revision stamp and its own .gitignore; the fix job adds scratch clones under .claude-security-run/patch-<timestamp>/, removed once each wave settles, and writes F<n>.patch, F<n>.md, PATCHES.md and patches.jsonl
- Network: role.md says a scan makes no network calls apart from the scan-changes pull-request search, which runs gh pr list only when a gh grant is present and gh is signed in, and fetches a pull request's head branch only after the user approves that network call
- Hooks (hooks.json): a banner when the /claude-security menu opens, usage counts after the plugin's Python helpers run, closing the scan's start question if it is unanswered for 60 seconds, and a scan suggestion after git push or gh pr create; usage counts go through Claude Code's telemetry unless DISABLE_TELEMETRY=1, DO_NOT_TRACK=1 or CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC=1 is set
- Hook files and commands: the banner and tip hooks write scan-tip.json in the plugin's data directory ($CLAUDE_PLUGIN_DATA); after a push or a new pull request the tip hook runs git rev-parse, symbolic-ref, merge-base and diff in the working directory with repository hooks and fsmonitor turned off, and it is skipped in CI or when CLAUDE_SECURITY_SCAN_TIP=off

### Cost model

Proprietary licence under the user's Anthropic terms; scans consume the session's model usage, and each scan asks once to confirm its time and token cost unless the request already accepts it

04 / FOLLOW THE EVIDENCE

## The source trail.

Our notes are separate from the original resource. Check upstream before adopting a new version.

 - [claude-security SKILL.md ↗](https://raw.githubusercontent.com/anthropics/claude-plugins-official/41f32e407b450055044f167db7ded5e8ff67d203/plugins/claude-security/skills/claude-security/SKILL.md) Checked 2026-10-21 https://raw.githubusercontent.com/anthropics/claude-plugins-official/41f32e407b450055044f167db7ded5e8ff67d203/plugins/claude-security/skills/claude-security/SKILL.md Supports: upstreamDescription, summary, whySelected[0], bestFor[0], bestFor[1], bestFor[2], review.findings[0], access.permissions[0], access.permissions[1]
- [claude-security role.md ↗](https://raw.githubusercontent.com/anthropics/claude-plugins-official/41f32e407b450055044f167db7ded5e8ff67d203/plugins/claude-security/skills/claude-security/role.md) Checked 2026-10-21 https://raw.githubusercontent.com/anthropics/claude-plugins-official/41f32e407b450055044f167db7ded5e8ff67d203/plugins/claude-security/skills/claude-security/role.md Supports: whySelected[0], review.findings[0], review.findings[1], review.findings[2], access.permissions[1], access.permissions[5]
- [claude-security jobs/scan-changes.md, pull-request search ↗](https://raw.githubusercontent.com/anthropics/claude-plugins-official/41f32e407b450055044f167db7ded5e8ff67d203/plugins/claude-security/skills/claude-security/jobs/scan-changes.md) Checked 2026-10-21 https://raw.githubusercontent.com/anthropics/claude-plugins-official/41f32e407b450055044f167db7ded5e8ff67d203/plugins/claude-security/skills/claude-security/jobs/scan-changes.md Supports: access.permissions[5]
- [claude-security jobs/suggest-patches.md ↗](https://raw.githubusercontent.com/anthropics/claude-plugins-official/41f32e407b450055044f167db7ded5e8ff67d203/plugins/claude-security/skills/claude-security/jobs/suggest-patches.md) Checked 2026-10-22 https://raw.githubusercontent.com/anthropics/claude-plugins-official/41f32e407b450055044f167db7ded5e8ff67d203/plugins/claude-security/skills/claude-security/jobs/suggest-patches.md Supports: limitations[8], access.permissions[2], access.permissions[4], whySelected[2]
- [claude-security agents/scan-researcher.md ↗](https://raw.githubusercontent.com/anthropics/claude-plugins-official/41f32e407b450055044f167db7ded5e8ff67d203/plugins/claude-security/agents/scan-researcher.md) Checked 2026-10-22 https://raw.githubusercontent.com/anthropics/claude-plugins-official/41f32e407b450055044f167db7ded5e8ff67d203/plugins/claude-security/agents/scan-researcher.md Supports: review.findings[5], access.permissions[2], install.steps[0]
- [claude-security agents/patch-generator.md ↗](https://raw.githubusercontent.com/anthropics/claude-plugins-official/41f32e407b450055044f167db7ded5e8ff67d203/plugins/claude-security/agents/patch-generator.md) Checked 2026-10-22 https://raw.githubusercontent.com/anthropics/claude-plugins-official/41f32e407b450055044f167db7ded5e8ff67d203/plugins/claude-security/agents/patch-generator.md Supports: review.findings[5], access.permissions[2], install.steps[0]
- [claude-security agents/patch-verifier.md ↗](https://raw.githubusercontent.com/anthropics/claude-plugins-official/41f32e407b450055044f167db7ded5e8ff67d203/plugins/claude-security/agents/patch-verifier.md) Checked 2026-10-22 https://raw.githubusercontent.com/anthropics/claude-plugins-official/41f32e407b450055044f167db7ded5e8ff67d203/plugins/claude-security/agents/patch-verifier.md Supports: review.findings[5], access.permissions[2], limitations[8], install.steps[0]
- [claude-security agents/claude-security.md ↗](https://raw.githubusercontent.com/anthropics/claude-plugins-official/41f32e407b450055044f167db7ded5e8ff67d203/plugins/claude-security/agents/claude-security.md) Checked 2026-10-22 https://raw.githubusercontent.com/anthropics/claude-plugins-official/41f32e407b450055044f167db7ded5e8ff67d203/plugins/claude-security/agents/claude-security.md Supports: install.steps[0], access.permissions[3]
- [claude-security agents/scan-verifier.md ↗](https://raw.githubusercontent.com/anthropics/claude-plugins-official/41f32e407b450055044f167db7ded5e8ff67d203/plugins/claude-security/agents/scan-verifier.md) Checked 2026-10-22 https://raw.githubusercontent.com/anthropics/claude-plugins-official/41f32e407b450055044f167db7ded5e8ff67d203/plugins/claude-security/agents/scan-verifier.md Supports: install.steps[0], access.permissions[2], review.findings[5]
- [claude-security agents/explore.md ↗](https://raw.githubusercontent.com/anthropics/claude-plugins-official/41f32e407b450055044f167db7ded5e8ff67d203/plugins/claude-security/agents/explore.md) Checked 2026-10-22 https://raw.githubusercontent.com/anthropics/claude-plugins-official/41f32e407b450055044f167db7ded5e8ff67d203/plugins/claude-security/agents/explore.md Supports: install.steps[0], access.permissions[2], review.findings[5]
- [claude-security agents/scan-inventory.md ↗](https://raw.githubusercontent.com/anthropics/claude-plugins-official/41f32e407b450055044f167db7ded5e8ff67d203/plugins/claude-security/agents/scan-inventory.md) Checked 2026-10-22 https://raw.githubusercontent.com/anthropics/claude-plugins-official/41f32e407b450055044f167db7ded5e8ff67d203/plugins/claude-security/agents/scan-inventory.md Supports: install.steps[0], access.permissions[2]
- [claude-security agents/scan-loader.md ↗](https://raw.githubusercontent.com/anthropics/claude-plugins-official/41f32e407b450055044f167db7ded5e8ff67d203/plugins/claude-security/agents/scan-loader.md) Checked 2026-10-22 https://raw.githubusercontent.com/anthropics/claude-plugins-official/41f32e407b450055044f167db7ded5e8ff67d203/plugins/claude-security/agents/scan-loader.md Supports: install.steps[0], access.permissions[2]
- [claude-security agents/scan-redactor.md ↗](https://raw.githubusercontent.com/anthropics/claude-plugins-official/41f32e407b450055044f167db7ded5e8ff67d203/plugins/claude-security/agents/scan-redactor.md) Checked 2026-10-22 https://raw.githubusercontent.com/anthropics/claude-plugins-official/41f32e407b450055044f167db7ded5e8ff67d203/plugins/claude-security/agents/scan-redactor.md Supports: install.steps[0], access.permissions[2]
- [claude-security README ↗](https://raw.githubusercontent.com/anthropics/claude-plugins-official/41f32e407b450055044f167db7ded5e8ff67d203/plugins/claude-security/README.md) Checked 2026-10-21 https://raw.githubusercontent.com/anthropics/claude-plugins-official/41f32e407b450055044f167db7ded5e8ff67d203/plugins/claude-security/README.md Supports: install.command, install.sourceUrl, install.steps[0], install.steps[1], install.steps[2], install.prerequisites[0], install.prerequisites[1], install.prerequisites[2], whySelected[0], whySelected[1], whySelected[2], bestFor[1], bestFor[2], bestFor[3], limitations[2], limitations[3], limitations[4], limitations[5], limitations[6], limitations[7], review.findings[1], access.permissions[4], access.permissions[6], access.label, access.cost, compatibility
- [claude-security SECURITY.md ↗](https://raw.githubusercontent.com/anthropics/claude-plugins-official/41f32e407b450055044f167db7ded5e8ff67d203/plugins/claude-security/SECURITY.md) Checked 2026-10-21 https://raw.githubusercontent.com/anthropics/claude-plugins-official/41f32e407b450055044f167db7ded5e8ff67d203/plugins/claude-security/SECURITY.md Supports: limitations[2]
- [claude-security LICENSE ↗](https://raw.githubusercontent.com/anthropics/claude-plugins-official/41f32e407b450055044f167db7ded5e8ff67d203/plugins/claude-security/LICENSE) Checked 2026-10-21 https://raw.githubusercontent.com/anthropics/claude-plugins-official/41f32e407b450055044f167db7ded5e8ff67d203/plugins/claude-security/LICENSE Supports: license, limitations[0], access.cost, compatibility
- [claude-security hooks/hooks.json ↗](https://raw.githubusercontent.com/anthropics/claude-plugins-official/41f32e407b450055044f167db7ded5e8ff67d203/plugins/claude-security/hooks/hooks.json) Checked 2026-10-21 https://raw.githubusercontent.com/anthropics/claude-plugins-official/41f32e407b450055044f167db7ded5e8ff67d203/plugins/claude-security/hooks/hooks.json Supports: limitations[6], access.permissions[6]
- [claude-security hooks/hooks.py ↗](https://raw.githubusercontent.com/anthropics/claude-plugins-official/41f32e407b450055044f167db7ded5e8ff67d203/plugins/claude-security/hooks/hooks.py) Checked 2026-10-22 https://raw.githubusercontent.com/anthropics/claude-plugins-official/41f32e407b450055044f167db7ded5e8ff67d203/plugins/claude-security/hooks/hooks.py Supports: review.findings[3], access.permissions[7]
- [claude-security hooks/hooks.sh ↗](https://raw.githubusercontent.com/anthropics/claude-plugins-official/41f32e407b450055044f167db7ded5e8ff67d203/plugins/claude-security/hooks/hooks.sh) Checked 2026-10-22 https://raw.githubusercontent.com/anthropics/claude-plugins-official/41f32e407b450055044f167db7ded5e8ff67d203/plugins/claude-security/hooks/hooks.sh Supports: access.permissions[7]
- [claude-security NOTICE.md ↗](https://raw.githubusercontent.com/anthropics/claude-plugins-official/41f32e407b450055044f167db7ded5e8ff67d203/plugins/claude-security/NOTICE.md) Checked 2026-10-21 https://raw.githubusercontent.com/anthropics/claude-plugins-official/41f32e407b450055044f167db7ded5e8ff67d203/plugins/claude-security/NOTICE.md Supports: review.findings[4]
- [claude-security plugin.json ↗](https://raw.githubusercontent.com/anthropics/claude-plugins-official/41f32e407b450055044f167db7ded5e8ff67d203/plugins/claude-security/.claude-plugin/plugin.json) Checked 2026-10-21 https://raw.githubusercontent.com/anthropics/claude-plugins-official/41f32e407b450055044f167db7ded5e8ff67d203/plugins/claude-security/.claude-plugin/plugin.json Supports: license

KEEP COMPARING

## Other approaches to consider.

Related by category or shared topics. These are alternatives to inspect, not a measured quality order.

 [### Trailmark Finding Triage ↗ Triages a single finding, SARIF result, weAudit annotation, suspicious function or report excerpt with Trailmark reachability, entrypoint-path, taint, privilege-boundary, blast-radius and caller/callee evidence, returning one of four verdicts: Promote, Needs manual review, Deprioritize or Blocked.](/skills/trailofbits-trailmark-finding-triage/)[### Vulnerability Triage Brocards ↗ Runs an incoming vulnerability report or automated finding through seven pass, dismiss or needs-more-info tests and returns a one-table triage summary with an accept, dismiss or needs-more-info verdict.](/skills/trailofbits-vulnerability-triage-brocards/)[### Audit Augmentation ↗ Overlays external findings — SARIF static-analysis results, weAudit annotation files and Trailmark 0.4.x binary-analysis exports — onto a Trailmark code graph by file and line overlap and builds severity-based subgraphs.](/skills/trailofbits-trailmark-audit-augmentation/)

 [AI Tools ↗](/tools/)[Skills ↗](/skills/)[Agents ↗](/agents/)[MCP Servers ↗](/mcp-servers/)[Workflows ↗](/workflows/)

## Continue your investigation

 - [Choose a toolkit](/tools/)
- [Inspect execution hosts](/agents/)
- [Check tool access](/mcp-servers/)
- [Check project dependencies](/check-repo/)
