Deep Research: OSS Bug Scanner — find unknown bugs via semantic analysis

[EduardPetraeus]

Summary

New analyzer module for codebase-archaeologist that finds logical and security bugs via semantic code understanding — reasoning about intent vs implementation rather than pattern matching.

Scope

• Python-first, targeting auth and data-leak class bugs initially
• Extension of existing analysis pipeline (new analyzer, not separate repo)
• CLI integration: codebase-archaeologist scan --bugs <path>

Detectors (v0.1)

1. Auth bug detector — missing auth checks, privilege escalation, token handling errors
2. Data leak detector — logged secrets, unmasked PII, overly broad serialization

Success Criteria

• Validated against 5+ known CVEs
• At least 1 novel bug found and responsibly disclosed
• False positive rate < 30% on 10 popular Python repos

Links

• BRIEF: ~/deep-research/ideas/oss-bug-scanner/BRIEF.md
• Parent: codebase-archaeologist semantic analysis pipeline

Estimated Effort

3-5 weeks to v0.1.