AI code security glossary
ANSWER FIRST
Short, plain definitions of the terms used across this guide, from agentic remediation to vibe coding. Where a term comes from a standard or a regulator, the definition links to that source.
Editorial assessment · Compiled from vendor product, docs, pricing and marketplace pages · Rated October 2026
- Active committer
- The unit GitHub uses to price GitHub Secret Protection and GitHub Code Security, at $19 and $30 per active committer per month. Check GitHub's billing documentation for how committers are counted. Read more about Active committer
Source: GitHub Advanced Security
- Agentic AppSec
- Application security practices and tools that account for AI agents writing and changing code, and that may use agents to find and fix issues.
- Agentic pentesting
- Penetration testing carried out by AI agents that probe an application or an AI system for exploitable issues. OX Security's Agentic Pentester and Snyk Evo Continuous Offensive Security are examples.
- Agentic remediation
- The use of AI agents to fix security findings, typically by writing a code change and checking that the issue is gone. Legit Security, Snyk, GitHub and Cycode each describe fix agents. Read more about Agentic remediation
- AI coding assistant
- A tool that suggests or writes code from natural-language prompts inside an editor or terminal, such as GitHub Copilot, Cursor or Claude Code. When it can also run commands and edit files on its own, this site calls it a coding agent.
- AI SAST
- Static analysis that uses AI models, alone or with rules, to find vulnerabilities and reduce false positives.
- AI-BOM
- An inventory of AI components in software and its development environment, such as models, AI packages, rule files, MCP servers and AI API keys. Read more about AI-BOM
- Anti-tampering
- Protection that stops a security control on a developer machine from being switched off or altered. Legit VibeGuard 2.0 lists anti-tampering among its endpoint features.
- ASPM
- Application security posture management; platforms that combine findings from many scanners, de-duplicate and rank them. Read more about ASPM
- CodeQL
- GitHub's code analysis engine, used by GitHub code scanning to find vulnerabilities. The CodeQL CLI runs the same analysis outside GitHub.
- Coding agent
- An AI assistant that can take actions such as editing files, running commands and calling tools, not only suggest code. Examples include Claude Code, Cursor's agent and GitHub's Copilot coding agent.
- Coding agent hook
- A user-defined command that runs at a point in an agent's lifecycle, such as before a tool call, and can block the action. Claude Code's PreToolUse hook is an example.
Source: Claude Code hooks documentation
- Container scanning
- Checking container images for known vulnerabilities in their operating system packages and application dependencies.
- Contributing developer
- A pricing unit that counts people who committed code in a recent window. Endor Labs counts contributors with commits in the last 90 days, and OX Security defines a developer as someone who committed in the past 90 days. Read more about Contributing developer
- Copilot Autofix
- A GitHub Code Security feature that suggests fixes for code scanning alerts.
- Cyber Resilience Act (CRA)
- EU regulation on the security of products with digital elements. From 11 September 2026, manufacturers must report actively exploited vulnerabilities and severe incidents through the CRA Single Reporting Platform, starting with an early warning within 24 hours. Read more about Cyber Resilience Act (CRA)
- DAST
- Dynamic application security testing: testing a running application from the outside, by sending requests and observing the responses, rather than reading its source code.
- Dependency cooldown
- A policy that holds back newly published package versions for a set period before they can be installed, so that malicious releases have time to be found and removed.
- IaC scanning
- Checking infrastructure as code files, such as Terraform templates or Kubernetes manifests, for insecure settings before they are applied.
- Install scripts
- Commands a package runs during installation, such as npm's preinstall, install and postinstall scripts. GitHub announced that npm v12 will not run them by default unless they are approved with npm approve-scripts. Read more about Install scripts
Source: GitHub Changelog: upcoming breaking changes for npm v12
- Malicious package
- A package published to a public registry with code meant to harm the machine or pipeline that installs it, for example by stealing credentials. Package firewalls and install-time checks aim to block them.
- MCP (Model Context Protocol)
- An open protocol that lets AI assistants connect to external tools and data sources through MCP servers.
- MCP security
- Controls on which MCP servers an agent may use and what data and actions they expose. Read more about MCP security
- Package firewall
- A control point that checks dependencies at install time and blocks malicious or disallowed packages.
- Pre-commit check
- A check that runs on a developer's machine before a commit is created, so that a secret or a known-bad pattern never enters Git history. Endor Labs describes pre-commit secrets detection and Apiiro lists pre-commit prevention. Read more about Pre-commit check
- Prompt injection
- Instructions hidden in content an AI model reads, such as code comments or dependency files, that try to change what it does.
- Push protection
- A check that blocks a git push when the commits contain secrets. GitHub Secret Protection includes it.
- Reachability analysis
- Determining whether vulnerable code in a dependency or function is actually called by the application, so unreachable findings can be deprioritized.
- Rule files
- Instruction files an AI assistant reads for persistent guidance, such as .github/copilot-instructions.md for Copilot or .cursor/rules for Cursor.
- SAST
- Static application security testing; analysis of source code for vulnerabilities without running it.
- SBOM
- Software bill of materials. CISA describes an SBOM as a nested inventory, a list of ingredients that make up software components.
Source: CISA: Software Bill of Materials
- SCA
- Software composition analysis; identification of open source dependencies and their known vulnerabilities and licenses.
- Secret scanning
- Detection of credentials such as API keys and tokens in code, Git history and other systems. Push protection and pull request merge rules act on its results. Read more about Secret scanning
- Secure prompt enrichment
- Adding security requirements or context to a developer's prompt before a coding assistant acts on it. Apiiro Guardian Agent lists secure prompt enrichment, and OX VibeSec sends security guidance to the agent before code is generated.
- Shadow AI
- AI tools, models or MCP servers in use without the security team's knowledge or approval. AI-BOM and inventory features aim to find them.
- Software supply chain security
- Protection of what goes into building and shipping software beyond first-party code: dependencies, build pipelines, CI/CD systems, registries and artifacts.
- SSDF
- Secure Software Development Framework, NIST SP 800-218: a set of secure software development practices in four groups, Prepare the Organization, Protect the Software, Produce Well-Secured Software and Respond to Vulnerabilities. NIST SP 800-218A adds a profile for generative AI. Read more about SSDF
- Vibe coding
- Building software mainly by prompting an AI assistant and accepting its output with little line-by-line review. Read more about Vibe coding