Endor Sast

Endor Labs SAST Scanner

Static application security testing for code-level vulnerabilities.

Vulnerability Categories

AI False Positive Reduction

Endor Labs offers AI-powered false positive filtering (requires Code Pro license). Before scanning, ask user if they want to enable it:

• Without AI (default): Faster, may include false positives
• With AI (Code Pro): Slower, filters false positives via AI review

Enable with --ai-sast-analysis=agent-fallback flag. If licensing error occurs, explain Code Pro requirement.

Workflow

Step 1: Run SAST Scan

Use scan MCP tool: scan_types: ["sast"], scan_options: { "quick_scan": true }.

The MCP scan tool accepts structured JSON only (path, scan_types, scan_options, optional namespace); it does not pass through arbitrary endorctl scan CLI flags such as --ai-sast-analysis. Use the CLI fallback below for AI false-positive reduction.

CLI fallback:

# Standard
npx -y endorctl scan --path $(pwd) --sast --output-type summary 2>/dev/null

# With AI false positive reduction
npx -y endorctl scan --path $(pwd) --sast --ai-sast-analysis=agent-fallback --output-type summary 2>/dev/null

Step 2: Retrieve Details

Use get_resource (resource_type: Finding) for each finding UUID from scan results.

Step 3: Analyze Code Context

Read source files referenced in findings. Show vulnerable code with surrounding context using file path and line numbers.

Step 4: Present Results

## SAST Analysis Results

**Path:** {path} | **Issues:** {count} | **AI FP Reduction:** {Enabled/Disabled}

### Critical Issues

#### {Issue #1}: {Title} ({CWE-ID})
**File:** {path}:{line} | **Severity:** Critical

**Vulnerable Code:**
```{lang}
{code snippet with line numbers}

Why dangerous: {brief explanation}

Fix:

{fixed code}

Summary

Next Steps

1. Fix critical issues first
2. /endor-sast — Verify fixes
3. /endor-scan-full — Full analysis
4. /endor-review — Pre-PR check

## Language-Specific Secure Patterns

**JS/TS:** `===` not `==`; avoid `eval()`, `Function()`, `setTimeout(string)`; `textContent` not `innerHTML`; `crypto.randomUUID()` not `Math.random()`

**Python:** parameterized queries not f-strings; `subprocess.run([], shell=False)` not `os.system()`; `secrets` module for crypto; avoid `pickle.loads()` on untrusted data

**Go:** `html/template` not `text/template`; `crypto/rand` not `math/rand`; `filepath.Clean()` for paths

**Java:** `PreparedStatement` for SQL; `SecureRandom` for randomness; avoid `Runtime.exec()` with user input

For data source policy, read references/data-sources.md.

## Error Handling

| Error | Action |
|-------|--------|
| No issues found | Confirm scan completed; suggest `/endor-scan-full` for deeper analysis |
| Auth error | Suggest `/endor-setup` |
| Unsupported language | List supported languages and alternatives |