leakguard

Fast, lightweight secret scanner for codebases (Rust CLI + Python package).

leakguard scans source files for accidentally committed credentials, tokens, and sensitive config values.

Why leakguard

• 104 built-in rules across cloud, LLM, database, HTTP auth, observability, and SaaS ecosystems
• Fast local scanning with sensible defaults
• CI-friendly outputs: pretty, json, sarif, markdown
• GitHub Actions summary support via --github-summary
• False-positive controls: inline ignore markers + rule-level disable in config
• Binary-safe scanning: non-text files are skipped
• Safe defaults: .env files are excluded

Quickstart

Install from source

git clone https://github.com/adrian-lorenz/leakguard.git
cd leakguard
cargo install --path .

Install via pip

pip install leakguard-secret-leaks
leakguard check .

Use prebuilt binaries

Download the latest release from GitHub Releases.

chmod +x leakguard-linux-amd64
sudo mv leakguard-linux-amd64 /usr/local/bin/leakguard

CLI

Common commands

# Scan current directory
leakguard check

# Scan specific path
leakguard check --source ./src

# JSON for automation
leakguard check --format json

# SARIF for GitHub code scanning
leakguard check --format sarif

# Markdown report (useful for CI summaries)
leakguard check --format markdown

# Show warning-level findings in details
leakguard check --warnings

# Verbose file-level logging
leakguard check --verbose

# Max file size in KB (default: 1024)
leakguard check --max-size 512

# Use custom config file
leakguard check --config /path/to/leakguard.toml

# Write GitHub Actions job summary
leakguard check --github-summary

# Show all built-in rules
leakguard rules

# Create default config
leakguard init-config

Exit codes

Python API

Use leakguard as a library, for example before sending text to an LLM:

from leakguard import scan_text, scan_text_dict, replace_text

findings = scan_text("My API key is sk-proj-abc123xyz...")
for f in findings:
    print(f.rule_id, f.severity, f.secret, f.secret_hash)

# Disable specific rules for this call
findings = scan_text("...", disable_rules=["http-insecure-url"])

# Optional replacer for secret output
findings = scan_text("...", replace_secret_with="[MASKED]")

# Directly sanitize text before sending it to an LLM
safe_text, replaced_any = replace_text(
    "My API key is sk-proj-abc123xyz...",
    replacement="[MASKED]"
)
if replaced_any:
    print("Secrets were replaced before LLM call")

Finding fields: rule_id, description, severity, line_number, line, secret, secret_hash, tags

Pydantic (dict-ready)

from pydantic import BaseModel
from leakguard import scan_text_dict

class Finding(BaseModel):
    rule_id: str
    description: str
    severity: str
    line_number: int
    line: str
    secret: str
    secret_hash: str
    tags: list[str]

rows = scan_text_dict("My API key is sk-proj-abc123xyz...")
findings = [Finding.model_validate(row) for row in rows]

Configuration

Generate a config file:

leakguard init-config

Default leakguard.toml:

[scan]
# Empty means all files except .env and .git
extensions = []
exclude_paths = []
exclude_files = []

[rules]
# Example: disable = ["jwt-token", "http-insecure-url"]
disable = []

Auto-loading behavior:

• leakguard.toml in the current working directory is loaded automatically.

Suppression and false positives

Inline suppression markers on a line:

api_url = "http://internal-service/api"  # leakguard-ignore

Supported markers:

• # leakguard-ignore
• # noqa-secrets
• # nosec-secrets

Built-in false-positive filtering includes patterns such as:

• templated values ({DB_PASSWORD}, %(password)s)
• shell variable references ($DB_PASSWORD)
• attribute references (settings.DB_PASSWORD)
• localhost URLs (http://localhost:8080)

Detection coverage

Current built-in categories include:

• Cloud / VCS: AWS, GitHub/GitLab, Google, Stripe, Slack, NPM, Docker Hub
• LLM / AI: OpenAI, Anthropic, Cohere, Mistral, Hugging Face, Groq, Perplexity, xAI, and related env leaks
• Azure / M365: tenant/app credentials, storage/service keys, Graph, Teams webhooks
• Databases / BaaS: PostgreSQL/MySQL/Mongo/Redis/MSSQL/JDBC, Supabase
• Observability: Datadog, New Relic, Grafana, Honeycomb, OTLP-related patterns
• HTTP/Auth: Bearer/Basic headers, credentialed URLs
• Crypto / Generic: PEM keys, JWT-like tokens, high-entropy secret assignments

Use leakguard rules for the complete, authoritative list.

GitHub Actions

Example workflow job:

jobs:
  leakguard:
    name: leakguard secret scan
    runs-on: ubuntu-latest
    permissions:
      contents: read

    steps:
      - uses: actions/checkout@v4

      - name: Install leakguard
        run: |
          curl -sSfL \
            https://github.com/adrian-lorenz/leakguard/releases/latest/download/leakguard-linux-amd64 \
            -o /usr/local/bin/leakguard
          chmod +x /usr/local/bin/leakguard

      - name: Run scan
        run: leakguard check --format markdown >> "$GITHUB_STEP_SUMMARY"

Alternative install method:

- name: Install leakguard
  run: pip install leakguard-secret-leaks

License

MIT — Copyright (c) 2026 Adrian Lorenz a.lorenz@noa-x.de