Repository navigation
Sudeshi Code Review v1.2.0 — Security Hardening
This release focuses on making the existing security checks more reliable and strengthening safeguards around AI-assisted review. It does not add new detection rules.
What's new
Security detection and evidence
- Improved
hardcoded-secretdetection for credentialed database URLs and quoted or statically computed object-property keys. - Redacted all detected secret literals from the affected finding-evidence line.
- Improved
unsafe-linkdetection for JavaScript URLs in direct JSX string expressions and static template expressions. - Avoided resolving shadowed outer variables as if they were local values.
- Corrected JSX parsing for
.jsand.jsxsource files so the existing JSX-based rules can analyze JavaScript files containing JSX.
AI-context redaction
- Hardened redaction for recognized secret values in supported source patterns, including quoted and statically computed object-property keys.
- Expanded regression coverage for repeated occurrences and edge cases.
- Clarified the best-effort limitation: the tool cannot guarantee that every secret will be detected or removed before source context is sent to Sarvam AI.
CLI and package
- Expanded CLI help with usage, options, environment variables, and examples.
- Verified short and long help/version flags.
- Validated the published-package layout and clean production installation, including the runtime TypeScript dependency and npm CLI executable.
Validation
- 176 tests passed across 19 test files.
- Formatting, lint, and build checks passed.
- The v1.2.0 tarball installed successfully in a clean consumer project with production dependencies only.
- The installed CLI was verified against a sample repository, including Markdown and JSON report generation and redacted secret evidence.
- Deterministic scanning was verified without
SARVAM_API_KEY.
Install
npm install --global sudeshi-code-reviewRequires Node.js 24 or later.
Important limitations
Sudeshi Code Review is a focused static-analysis tool, not a comprehensive security scanner. Findings can include false positives and false negatives. AI-context redaction is best-effort and must not be treated as a guarantee that every secret is removed. Review the README and SECURITY.md before enabling AI review on private or sensitive source code.
Full Changelog: v1.1.1...v1.2.0