REPORT

Audit Report

Project: LazyOwn | 2026-10-17 | offline, deterministic

Confidence Trail

Every edge is tagged. Extracted means parsed from source; inferred means derived heuristically; ambiguous is reported, never hidden.

Confidence	Count	Meaning
EXTRACTED	4517	Resolved import edges parsed from source
EXTRACTED	10408	Raw import statements (may include externals)
INFERRED	5	Surprising cross-community bridges
AMBIGUOUS	0	No uncertain edges are emitted by the static scanner

Coverage

• Files: 1078, communities: 23
• File doc coverage: 949/1078
• Orphans (no docs at any level): 103
• Layers detected: 6
• Security findings: 0
• Large files (>256KB, maybe generated): 9 (lazyc2.py, lazy_loop.py, lazyown_mcp.py, mcp_generated_tools.py, 3d-force-graph-1.73.4.min.js)

Limits

• Python uses the ast module; all other languages use regex parsers.
• No dataflow or runtime tracing; taint follows the import graph only.
• Symbol docs come from adjacent comments; missing docs are listed, not invented.
• Centrality scores count every scanned file equally, including checked-in build artifacts; verify large files before refactoring.

Token Benchmark

• Wiki index plus community pages estimate: ~791985 tokens (chars/4).
• Full re-read of every source file would cost strictly more on any non-trivial project; this wiki is the cheaper entry point.
• Generation cost: $0, offline, no network calls.

Reproduce

readmenator . --rebuild
readmenator . wiki