SCAN. DETECT. REPORT.
An AI smart contract security auditor with four product modes. 28 vulnerability categories. Multi-model consensus in seconds, not days.
$ prion scan contract.sol --mode aconit [ACONIT mode engaged] [scanning 28 vulnerability categories] > REENTRANCY ................ clean > OVERFLOW .................. clean > ACCESS_CONTROL ............ clean > ORACLE_MANIPULATION ....... finding · HIGH > FLASH_LOAN_ACCOUNTING ..... clean [consensus] 0 critical · 1 high · 2 info [report] ./report_3b4a.json $
ACONIT.SCAN
ACONIT—SCAN.
Comprehensive audit mode. Full dependency tracking, cross-contract state analysis, invariant checking across all 28 vulnerability categories. The deepest scan in the toolkit — built for pre-deployment and bounty prep.
$ prion scan contract.sol --mode aconit --depth full [aconit loaded · cross-contract enabled] [phase 1/6] static analysis ................ done [phase 2/6] screening ...................... 3 flagged [phase 3/6] synthesis ...................... 3 candidates [phase 4/6] consensus ...................... 2/3 verified [phase 5/6] invariants ..................... 1 violation [phase 6/6] PoC replay ..................... confirmed [REPORT] aconit_report.json · 2 critical · 1 high · 4 info $
ABRIN.CI
ABRIN—SCAN.
The CI/CD-friendly mode. Eight detection heads in parallel, consensus in seconds. Drop it into your pipeline — every commit scanned, every pull request validated before merge.
# .github/workflows/prion.yml name: PRION Rapid Scan on: [pull_request] jobs: audit: runs-on: ubuntu-latest steps: - uses: actions/checkout@v4 - name: Abrin Scan run: prion scan ./contracts --mode abrin --format junit - name: Publish findings uses: prion/gh-annotate@v1
NANGA.RUN
NANGA—MARK.
The regression framework. Runs against the DeFiHackLabs dataset — 699 documented mainnet security incidents — to validate detection rules and measure recall across model versions.
ERGOT.LOOP
ERGOT—ENGINE.
Randomized input fuzzing and mutation-based synthesis. Finds rare edge cases — unusual state transitions, unlikely input combinations, subtle accounting drift. Runs continuously in the background while you code.
28 CATEGORIES. ZERO BLIND SPOTS.
HUNTER DETECTS. VERIFIER CONFIRMS.
HUNTER
Generates hypotheses. Writes PoCs. Explores unusual patterns across contract state and call graphs.
- Static + symbolic analysis pre-pass
- Pattern library across 28 categories
- Generates candidate exploits as PoC tests
- Forks mainnet state for realistic replays
VERIFIER
Challenges each hypothesis. Rejects plausible-looking noise. Survivors become real findings.
- Adversarial re-reading of the PoC
- Invariant checks against the hypothesis
- Confidence scoring + consensus vote
- Outputs reproducible report with severity
THREE WAYS TO WORK.
CI/CD MONITORING
Rapid mode on every commit. JUnit reports in your PR annotations. Weekly deep-scan sweep of main.
CUSTOM AUDIT
Deep full-depth pass + PoCs + remediation. Direct-to-engineer Slack or Telegram channel during the audit.
BOUNTY SUCCESS FEE
Run against active Immunefi programs. No fee unless a finding is accepted and paid out by the program.
SUBMIT
FOR AUDIT.
Solo-operated AI auditor. No sales funnel, no account managers. Your contract lands in front of the engineer running the models within minutes.