Browse the directory

Source-backed rules for reviewing AI-generated database access code for SQL injection before merge, covering parameterized queries, identifier handling, ORM safety, dynamic query construction, least-privilege access, and privacy-safe test evidence.

Set up Claude Code in GitLab CI/CD with maintainer-approved prompts, least-privilege tool access, and Bedrock or Vertex OIDC alternatives from official GitLab integration docs.

Secure Claude Code hooks in shared repositories: version-control hook configs, code review for PreToolUse and PostToolUse scripts, least-privilege matchers, secret scanning, and rollback when hook behavior changes.

A practical walkthrough of least-privilege permission design in the Claude Agent SDK. Learn the evaluation order (hooks → deny → mode → allow → canUseTool), how allowedTools and disallowedTools differ, the six permission modes, and how to lock an agent to a fixed tool surface with dontAsk.

A source-backed guide for designing Model Context Protocol servers with explicit authorization boundaries, narrow tools, scoped resources, privacy-aware logging, and least-privilege runtime access.

Source-backed rules for safely connecting MCP servers that expose local tools, resources, prompts, roots, transports, credentials, files, and logs to AI clients during active development sessions.

A practical pre-installation review workflow for Model Context Protocol servers. Inventory tools, resources, prompts, credentials, network reach, storage, and runtime permissions before connecting a server to Claude or another MCP client.

Apache-2.0 security scanner from NVIDIA for AI agent skills, with static pattern checks, optional LLM semantic analysis, MCP least-privilege and tool poisoning analyzers, OSV.dev vulnerability lookups, risk scoring, and terminal, JSON, Markdown, and SARIF reports.

Expert MCP OAuth server hardening capability pack applying documented Dynamic Client Registration, oauth.scopes pins, callback ports, keychain token storage, and least-privilege scope review from official MCP documentation.

Source-backed rules for reviewing AI-generated code that builds or runs operating-system commands, shell invocations, or subprocesses before merge for command injection and argument injection risk, covering library alternatives to shelling out, array-form process APIs, allowlist input validation, and least-privilege execution.

Set up Claude Code GitHub Actions for pull request review: install the Claude GitHub app, store ANTHROPIC_API_KEY in secrets, pin anthropics/claude-code-action to an audited commit SHA with least-privilege permissions, and follow documented security practices.

Defend tool-connected agents against prompt injection using documented Claude Code security practices: MCP trust verification, approval gates, least-privilege tools, untrusted content handling, and human review before side-effect tool calls.

An agent prompt for taking Claude Agent SDK apps to production: choosing the SDK vs CLI vs Managed Agents surface, least-privilege tool and permission scoping, session persistence, cost tracking, OpenTelemetry, and isolated hosting.

Source-backed specialist agent for reviewing remote MCP authorization boundaries, protected resource metadata, resource indicators, token audience validation, token passthrough risk, and least-privilege scopes.

Source-backed collection for reviewing OAuth-backed and remote MCP servers: protected resource metadata, token audience checks, least-privilege scopes, config privacy, local tool access, and interactive server inspection.

Source-backed rules for reviewing remote MCP server authorization boundaries: protected resource metadata, OAuth resource indicators, token audience checks, least-privilege scopes, and cross-server token isolation.

Source-backed agent that threat-models an MCP server before it is connected to Claude Code, covering trust verification, tool authority and side effects, prompt injection via tool output, network and credential exposure, and least-privilege mitigations, grounded in the official security docs.