harvest-scan
Fast, plugin-driven security & code-quality scanner. One command. One score. Ten plugins. Zero config.
Actionable feedback, not a wall of noise
Grouped by file, colour-coded by severity, with a score bar and a trend since your last run.
Three tools' worth of signal, one number.
Most teams juggle a linter, a dependency auditor, and a SAST scanner. harvest-scan runs them together and answers one question: is this code better or worse than last time?
One number to watch
Every scan yields a single 0–100 score and letter grade, so quality trends are obvious across commits and PRs.
Fast & parallel
Plugins run concurrently with read-once file caching; binaries and huge files are skipped automatically.
10 plugins built in
Critical patterns, heuristics, ESLint, complexity, AST security, Semgrep, npm audit, OSV CVEs, Dockerfile, and NLP.
Auto-discovered plugins
Drop a file in plugins/ (or run harvest plugin create) — it is picked up automatically, no wiring.
Autofix
--fix applies safe ESLint fixes before scanning, so you ship cleaner code without manual edits.
SARIF + JSON
Findings show up inline on GitHub PRs via SARIF; JSON output plugs into any tooling.
Baseline / ratchet
Adopt on legacy code and fail CI only on new issues — never drown in the existing backlog.
CI-native
Real exit codes and a ready-made GitHub Action. Gate merges on quality with one step.
A small orchestrator around composable plugins
Discover
Globs the target respecting .gitignore, skipping node_modules, binaries, and files over 512 KB.
Load plugins
Every plugin in the plugins directory is auto-discovered and filtered by scan mode (fast / default / complete).
Run in parallel
Active plugins run concurrently with per-plugin error isolation — one failure never aborts the scan.
Score
Issues are weighted into a 0–100 score. Any critical finding fails the scan outright.
Scoring model
Start at 100, subtract a weight per issue. A single critical finding fails outright.
| Score | Rating | Grade | Exit code |
|---|---|---|---|
| 90–100 | excellent | A | 0 |
| 75–89 | good | B | 0 |
| 50–74 | fair | C | 0 |
| 25–49 | poor | D | 1 |
| 0–24 | bad | F | 2 |
Ten checks, one score
Each plugin decides which files it applies to and returns structured issues. Add your own by dropping a file in plugins/.
Destructive / RCE patterns — rm -rf /, eval(), child_process.exec, curl | sh, fork bombs. Any hit fails the scan.
applies to: all files
Risky calls & secrets — eval, execSync, shell_exec, base64_decode, hard-coded credentials.
applies to: all files
Runs ESLint against your project and counts errors.
applies to: JS / TS
Functions exceeding the cyclomatic complexity threshold.
applies to: JS / TS
AST-level security analysis via @nodesecure/js-x-ray.
applies to: JS / TS
Image hygiene — non-root USER, pinned tags, leaked secrets, exec-form CMD.
applies to: Dockerfile*
OWASP Top Ten rules across 30+ languages.
applies to: many langs
npm audit severity rolled into the score.
applies to: package.json
Known CVEs for your dependencies via the OSV.dev API.
applies to: package.json
NLP pass flagging dangerous security terms in any text.
applies to: all files
Running in ten seconds
Requires Node.js 18+. Zero config to start.
harvest -ffast scan (critical + heuristics + lint)harvest src/ -ccomplete scan (every plugin)harvest . --fixautofix, then scanharvest . --sarif -o harvest.sarifSARIF for GitHub Code Scanningharvest baseline .accept current issues; fail only on new onesGitHub Action
- uses: actions/checkout@v4
- uses: Aditya060806/Harvest@v1
with:
target: '.'
mode: 'complete'
- uses: github/codeql-action/upload-sarif@v3
with:
sarif_file: harvest.sarif