v3.6: Respect .aiexclude/.geminiignore during repo discovery + batch reporting

[pauldatta]

Feature Requests for v3.6

1. Respect exclusion files during recursive repo discovery

Currently the scanner uses a hardcoded skip list (node_modules, .venv, dist, etc.) when walking directories for --repos. It should also respect:

• .aiexclude — Gemini CLI's native exclusion mechanism
• .geminiignore — Gemini CLI context exclusion
• .aiignore — Emerging convention across AI tools

Rationale: Users who have explicitly marked directories as off-limits for AI tools should have that respected during scanning. The hardcoded list covers ~90% of noise, but .aiexclude gives users precise control.

Scope: Parse glob patterns from these files at each directory level during the walk. Handle negation rules (!) and nested ignore files.

---

2. Batch report processing module

Scaffold a reports/ module that can:

• Ingest multiple gemini-env-manifest.json files from different team members or time periods
• Generate trend analysis (maturity score over time, tool adoption curves)
• Produce team dashboards aggregating insights across an organization
• Export to CSV, HTML, and PDF formats

Reference: reports/README.md already contains the planned directory structure.

---

3. Unit tests for new v3.5 advisory checks

The 11 new advisory checks (checkMemoryHygiene, checkSkillExtraction, expanded checkPolicyHygiene) need dedicated unit tests in test/advisor.test.js covering edge cases:

• Memory bloat threshold (exactly 5000 words)
• Stale lock boundary (exactly 30 minutes)
• YOLO mode with/without deny rules
• Cross-tier duplication detection accuracy

---

Labels

• enhancement
• v3.6