Semgrep
14,901Fast, rule-based static analysis with a pattern syntax that mirrors source code.
Maturity tier details
Computed tier: 2A(raw score 5.17). Effective tier: 2A.
Facets:documented_in=0evaluated_on=0integration_surfaces=1.5paper_citations=0recent_commit=2stars=1.67
Source of truth:pipeline/maturity/score.py
Description
Open-source static-analysis engine that matches syntactic and lightweight-dataflow patterns written in a source-code-like DSL, spanning 30+ languages. Ships a large community rule registry (OWASP Top 10, secrets, framework anti-patterns) and supports custom rules that can flag reinvented-wheel patterns - e.g., hand-rolled retry loops, ad-hoc string-escaping, or reimplemented stdlib utilities - surfacing places where an agent wrote code that an existing library call would have covered. Runs in CLI, pre-commit, and CI; Semgrep AppSec Platform adds hosted findings triage.
Why it's included
Occupies detection x post_generation x fully_autonomous for dependency_blindness: custom rules catch patterns such as "string built with manual concatenation where shlex.quote exists" or "home-rolled HTTP retry where tenacity is on the requirements", pointing reviewers at library-duplicating code after the fact. Distinct from tree-sitter (structural substrate, prevention) and sourcegraph-cody (retrieval grounding) because Semgrep is rule-based diagnosis on the finished diff, making it the classic CI-gate node in the MECE grid.
Security notes
Core engine is LGPL-2.1 and runs fully offline; the default rule registry is pulled from semgrep.dev and should be pinned by commit or vendored for regulated builds. The hosted AppSec Platform is SaaS and sends findings (not raw source by default) upstream; self-hosted deployments keep everything on-prem. Rules execute as pattern matching, not arbitrary code, so registry compromise is bounded to false-positive/negative impact rather than RCE.
Addresses failure modes
- Dependency blindness — Reimplementing functionality that already exists in reachable code.
- Security vulnerability — Insecure patterns, injection flaws, weak cryptography, secret leakage.