CodeGuard

Fast, offline static analysis for Python, JavaScript, and TypeScript. It finds security vulnerabilities (with source-to-sink taint tracking), performance problems, and code-quality risks, fixes the safe ones for you, and works at every gate of your workflow: editor, pre-commit, PR/CI, and scheduled audit, all from one config file.

Status: 2.2.0 (beta). Rule IDs, the Finding / JSON schema, config keys, and exit codes have been a stable contract since 2.0. Upgrading never changes a pipeline's exit code: scan and ci run security rules unless you opt into more. See the migration notes if you used the 0.1 alpha.

What it does

app.py:11:5  [CG-SEC-011] CRITICAL  Untrusted input reaches a shell command
    11 |     os.system(cmd)
             ^
  ↳ line 9: Untrusted input: flask request.form
  ↳ line 9: Flows through: host = request.form["host"]
  ↳ line 10: Flows through: cmd = "ping -c 1 " + host
  → Run the program with an argument list and shell=False ...

Every finding has a stable rule ID, a severity, a CWE / OWASP mapping for security rules, a precise location, and a fix suggestion. Many also carry a safe autofix.

Full list: docs/rules, or codeguard list-rules.

Install

pipx install codeguard-cli      # or: uv tool install codeguard-cli
pip install codeguard-cli       # into the current environment

The PyPI project is codeguard-cli; the command is codeguard. It needs Python 3.10+. No database, no network, and no code leaves the machine.

From source:

git clone https://github.com/mevichitra/codeguard && cd codeguard
pip install -e ".[dev]"

Usage

codeguard scan src/                                  # security rules (default)
codeguard scan src/ --category performance           # or quality, or several
codeguard scan src/ --fix-dry-run                    # preview safe autofixes as a diff
codeguard scan src/ --fix                            # apply them
codeguard scan src/ --format sarif -o results.sarif  # json, sarif, github, rdjson, junit
codeguard scan src/ --fail-on high                   # exit 1 only for high or critical
codeguard explain CG-SEC-011                         # what a rule catches and how to fix it

Exit codes: 0 clean, 1 findings at or above --fail-on, 2 usage error, 3 config error, 4 internal error.

Suppressions

query = f"SELECT * FROM t WHERE id = {uid}"  # codeguard: ignore[CG-SEC-001] reason: uid is an int until=2026-12-31

A suppression without a reason, or past its until= date, is itself reported (CG-META-001 / 002). Audit them with codeguard suppressions list.

Configuration

codeguard.toml (or [tool.codeguard] in pyproject.toml). Run codeguard init for a starter file.

[codeguard]
exclude    = ["tests/fixtures/**", "**/*.min.js"]
categories = ["security", "performance", "quality"]   # default: ["security"]
fail_on    = "high"
baseline   = ".codeguard-baseline.json"

[codeguard.rules]
disable = ["CG-QUAL-104"]

[codeguard.rules.CG-QUAL-003]
max_complexity = 20          # rule options: see `codeguard explain <ID>`

See configuration, baselines, autofix, and taint tracking.

CI integration

- uses: mevichitra/codeguard/action@v2    # diff-aware scan + SARIF upload to code scanning

codeguard ci scans only the files changed since the base branch and applies the baseline. Pre-commit hooks and a Docker image are also available. See docs/ci.md.

Editor integrations

The language server (codeguard lsp) and the extensions put every finding in the editor with severity-coloured diagnostics, hovers that show the taint path, quick fixes, one-click suppressions, and fix-all on save. Editors show security, performance, and quality findings by default; CI keeps the project's categories.

• VS Code (editors/vscode): a Findings sidebar grouped by severity, category, or file; live status-bar counts; a Security Overview panel; fix all in the workspace; SARIF export; and a getting-started walkthrough.
• Zed (editors/zed): native diagnostics and Project Diagnostics, lsp.codeguard settings passthrough, code_actions_on_format fix-all, and a local Markdown dashboard.
• Neovim, Helix, and other LSP clients: run codeguard lsp. See docs/editors.md for configuration and the custom codeguard/summary protocol.

Contributing

See CONTRIBUTING.md. A new rule is one module, a vulnerable and a safe fixture (plus a fixed fixture if it has an autofix), a test, and a docs page.

Roadmap

Phase 1 (2.2, this release) added taint tracking, the performance and quality packs, autofix, and the editor UX. Next come JavaScript / TypeScript taint tracking, same-file function summaries, Semgrep-compatible custom rules, more security rules (SSRF, XXE, ReDoS, JWT, insecure cookies), Marketplace / Open VSX publishing, and standalone binaries. Details in docs/roadmap.md.

License

Apache-2.0. See LICENSE.

Contributions are accepted under the Developer Certificate of Origin (DCO). Sign your commits with git commit -s.