Enhance all 45 skills per 7-rule audit evaluation

[kamalsrini]

Summary

• 167 files changed (+7,332 / -2,075 lines) across all 45 skills in 10 domains
• ~75 new supporting files created under /references/, /scripts/, /templates/ directories
• 112 mandatory actions executed from SKILL_AUDIT.md evaluation against 7 rules

What changed

R4 File Structure (28 skills fixed): Extracted inline content (regex patterns, CWE mappings, framework tables, report templates, detection patterns) into structured /references/, /scripts/, /templates/ directories across all domains

R2 Verification (33 skills fixed): Added falsifiable tests with Expected Behavior, Actual Behavior Check, and binary pass/fail test cases

R5 Gotchas (15 skills fixed): Added/strengthened gotchas sections with false positive patterns, precision traps, and exploit pattern lessons

R3 Elegance (3 major dedup):

• owasp-top-10-web <> secure-code-review: Added delegation note, extracted overlapping detection patterns
• agentic-top-10 <> agent-security: Defined clear boundaries (framework compliance vs architecture review)
• iam-review: Refactored as orchestrator delegating to access-review, privileged-access, zero-trust-assessment

R1 System Layer (6 skills fixed): Added executable Grep/Glob detection patterns to prose-only skills:

• llm-top-10: 57 new detection patterns across all 10 categories
• zero-trust-assessment: Grep patterns for all 5 ZT pillars
• prompt-injection, agentic-top-10: Concrete regex patterns added

R7 Subagent (12 skills fixed): Added parallelization markers for independent assessment steps

Also includes

• SKILL_TEMPLATE.md for generating future skills
• SKILL_AUDIT_TEMPLATE.md for future evaluations
• @unitoneai/skills npm package (npx @unitoneai/skills init)

Test plan

• Verify all 45 SKILL.md files retain valid YAML frontmatter
• Verify Prompt Injection Safety Notice preserved on all skills
• Spot-check 5 skills: extracted content matches original
• Verify no broken file references (all -> See references/ paths exist)
• Run npx @unitoneai/skills init against clean project