perf: skip lockfiles (package-lock.json, yarn.lock, etc.)

perf: reduce tool calls ~80%, single-pass Python scanner

perf: 16x faster bash scanner, Python now primary

docs: update README for Tier 1 (60 patterns, charset-aware entropy, LLM rewrite)

feat: rewrite Phase 3 LLM analysis with correlation, exec summary, tech stack detection

feat: add 16 secret patterns (DigitalOcean, Vault, Grafana, Shopify, Anthropic, etc.)

feat: charset-aware entropy detection (hex 3.5, base64 4.2, generic 4.5)

perf: optimize bash scanner file filtering and dangerous file scan

feat: add dangerous file type detection, secret data safety, CRLF fix

docs: replace ASCII art header with SVG for clean GitHub rendering

docs: improve examples and add framework-specific false positive guidance

refine: tighten skill description to improve trigger accuracy

feat: improve skill quality scores across all dimensions

chore: update author name to Kuoshih Yang

feat: add agentskills.io spec fields to SKILL.md frontmatter