You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Confirmed (TG-INPUT-004 - Raw unvalidated filename)
✅ PASS
SQLAlchemy
safe-bound-query/
Safe
No finding (TG-INPUT-002 - Parameterized :param)
✅ PASS
SQLAlchemy
missing-tenant-scope/
Vuln
Confirmed (TG-AUTH-007 - Unscoped tenant query)
✅ PASS
🏢 Authorized Real-World Repository Evaluation
Repo ID
Stack Profile
Commit Evaluated
Detection Accuracy
Confirmed Findings
False Positives
Remediation Validated
Repo-A
Django 4.2 + DRF 3.14 + PostgreSQL
d5a1f80 (Authorized SaaS Base)
100%
2 (TG-AUTH-006, TG-RATE-001)
0 (1 downgraded to Manual Review)
✅ Verified
Repo-B
FastAPI 0.109 + Pydantic v2 + HTTPX
b92ce41 (Authorized Microservice)
100%
2 (TG-SSRF-001, TG-AUTH-006)
0
✅ Verified
Repo-C
Flask 3.0 + SQLAlchemy 2.0 + Flask-WTF
e718bc3 (Authorized Portal Base)
100%
2 (TG-INPUT-002, TG-INPUT-004)
0
✅ Verified
🔎 Loophole and Drawback Analysis
1. Detection Loopholes
Identified Risk: Monorepos with multiple sub-packages might mask backend dependencies if only root files were checked.
v0.4.1 Resolution: Added sub-directory traversal rules in SKILL.md and verified with tests/fixtures/python/stack-detection/mixed-monorepo/.
2. Rule Loopholes
Identified Risk: False positives on domain-driven architectures where authorization is delegated from controllers to service layers.
v0.4.1 Resolution: Updated TG-AUTH-007 to require Manual Review classification when service layer lookups are present rather than issuing unwarranted Confirmed alerts.
v0.4.1 Resolution: Provided framework-idiomatic remediation alternatives across ViewSet querysets, custom DRF BasePermission classes, and Pydantic model configurations.
4. Human-Review Loopholes
Identified Risk: Vague warnings like "Check permissions" provide little value to developers.
v0.4.1 Resolution: Standardized exact, testable prompt questions in audit reports (e.g., "Does OrderService.get_for_user enforce tenant ownership before returning records?").
🛡️ Security, Workflows, and Supply Chain Verification
Secret Scan Audit: Verified 0 hardcoded production secrets, API keys, or private certificates across tracked files and git history. All environment templates use explicit .env.example placeholder values.
CI/CD Workflows: All 5 GitHub Actions workflows (python-fixtures.yml, docs-links.yml, dependency-review.yml, secret-scan.yml, release-check.yml) verified with pinned action commit SHAs and minimal permissions (contents: read).
Documentation Link Integrity: Verified all 19 key documentation files and internal Markdown relative links with zero broken links.
License Integrity: Standard MIT License verified in root LICENSE and accurately referenced in README.md.
🚦 Final Release Recommendation
Decision Gate
Status
Finding / Action
Release v0.4.1 Status
🚀 APPROVED
Quality patch is stabilized, regression-tested, and clean.
Breaking Changes
None
100% backward-compatible with v0.3.x and v0.4.0.
Outstanding Bugs
None
All false positives and stack detection edge cases addressed.
Next Roadmap Horizon
Scheduled
v0.5.0 (Serverless & Edge Compute Security for Cloudflare & Vercel).
⚖️ Limitations & Boundary Statement
This evaluation does not prove complete vulnerability coverage, 100% application security, or universal fitness for every custom proprietary runtime. TorusGuard is an open-source Markdown-first security guidance framework for AI coding agents and human developers, designed to provide structured guardrails, audit methodologies, and remediation patterns.