Skill Forge v3.3.0

What's New

Security Red Line Check

• 7-item security scan before delivery (curl/wget, credentials, sensitive paths, eval/exec, system files, obfuscated code, cookies)
• Do NOT scope added to description

Full Pipeline

• Phase 0: Intent recognition + Adaptive 2-5 round interview (B1-B6 methodology)
• Phase 1: Creation + Self-validation (Schema + Security + Trigger + Dogfood)
• Phase 2: SkillHub peer benchmarking (Tencent 9-dimension comparison)

Three Iron Rules

• Description-first: vague description = dead Skill
• One-Skill-One-Job: multi-purpose = chaotic triggering
• Under 150 lines: over 200 = quality decay

Progressive Disclosure

• SKILL.md kept under 170 lines
• Detailed methodology in references/ directory