MCP server for detecting accidentally committed secrets, API keys, tokens, and credentials in your codebase
Server has 9 tools with complete input schemas and descriptions. Naming follows verb_noun convention (scan_*, get_*, add_*, set_*, exclude_*, generate_*). Descriptions are present but brief (avg ~40 chars, below 194-char baseline). Parameters are typed with descriptions. However, output schemas are not documented, LLMs cannot predict return structure. Error handling is minimal; no recovery guidance. No tool annotations (readOnlyHint/destructiveHint). Security concern: add_custom_pattern and exclude_false_positive modify state but lack confirmation/dry-run patterns.
Add a custom detection pattern
Add a finding to the whitelist
Generate a formatted report from scan results
Get current detection configuration
Get remediation advice for a specific secret type
Scan directory recursively for secrets
Scan single file for secrets
Output schemas not documented. LLMs cannot predict return structure for any tool. Scan results, remediation advice, and report formats are opaque.
Destructive/write tools lack confirmation or dry-run patterns. add_custom_pattern and exclude_false_positive modify state without user confirmation, risking accidental configuration changes.
No tool annotations. Tools lack readOnlyHint/destructiveHint/idempotentHint metadata, preventing clients from applying safety policies or UI hints.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 54 | <=2025-11-25 | v2 |
Set entropy detection threshold
Validate if text matches known secret patterns
Descriptions are too brief (avg 40 chars vs 194-char baseline). get_detection_config has only 'Get current detection configuration', no context on when to call it or what it reveals.
Error handling is minimal. No recovery guidance, no actionable error messages, no categorization of retryable vs fatal errors. LLMs cannot self-correct on failures.