socrateslisted

Five questions per principle: 1. Still true? (conditions, paths, tools current?) 2. Justified from first principles? (real incident / system constraint, not convention) 3. Best practice or just "what we've always done"? 4. Still violated? (check logs, commits, feedback files - never-fires = zombie or success) 5. Zombie? (already covered by higher-tier enforcement - hook/guard/harness?) Verdict scheme: | Verdict | Meaning | Action | |-----------|------------------------------|------------------------------------| | JUSTIFIED | Survives all 5 questions | Keep as-is | | VERBOSE | Justified but over-stated | Simplify to core trigger form | | WEAK | Justified but doesn't fire | Strengthen - move to hook, sharpen | | ZOMBIE | Fails Q1, Q4, or Q5 | Delete | | DEAD-REF | Rule fine; cross-refs broken | Fix the refs | Output format: table of per-principle verdicts (scanable, not prose), then one-paragraph summary of totals. Every principle must receive a verdict - "nothing should be unexamined." Anti-pattern: emerging with 0 zombies = you summarised, not questioned. When NOT to use: designing new things (/aristotle), ranking (/prioritise), rules just written (give it 2 weeks before questioning). Synergy: wire to a recurring task (quarterly) that passes your enforced-rules.md + CLAUDE.md as arguments. The skill is single sourc