lizikk
e78e7b403e
feat: add khazix-writer skill and strengthen write-article with 4 techniques
...
Install khazix-writer skill (WeChat public account writing style by 卡兹克).
Borrow 4 techniques to strengthen write-article's quality system:
- Punctuation ban (§2.5): forbid colons, em dashes, double quotes; tech_blog exempted
- Colloquial phrase library (§2.3.1): 5 categories of informal expressions
- 3-layer final review (replacing old 二次审计): T1 punctuation + T2 rhythm + T3 posture
- HKR topic quality check (§4.5): Happy × Knowledge × Resonance framework
Quality gate: 27 → 28 items.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
2026-04-07 18:21:44 +08:00
lizikk
376bf52ccd
refactor: sync 6 project skills to atomstorm upstream via symlinks
...
Replace local skill directories with symlinks pointing to
atomstorm-claude-skills upstream (b83c4d9):
write-article, paper-illustration, my-world, dao-analysis,
import-draft, multi-review
Also removes stale multi-review and paper-illustration variants.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
2026-03-29 22:33:17 +08:00
lizikk
e12c995c3a
feat: add 4 atomstorm skill symlinks (context-engineer, long-task-manager, multi-agent-orchestrator, tool-designer)
...
Link project skills to upstream atomstorm-claude-skills repo for shared agent engineering capabilities.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
2026-03-29 21:29:09 +08:00
lizikk
bb92dc5dc9
feat: add 8 agent engineering skills (v1.0.0)
...
- agent-harness: validation infrastructure builder
- context-engineer: context optimization (40-90% cost savings)
- tool-designer: ACI-compliant tool design
- long-task-manager: crash recovery with dual-agent pattern
- multi-agent-orchestrator: JSONL protocol + Worktree isolation
- agent-evaluator: Transcript + Outcome dual-dimension evaluation
- agent-security-auditor: three-layer security model
- agent-architect: end-to-end system design guide
All skills pass Anthropic best practices audit (5/5 stars).
P1 optimization completed: enhanced descriptions with all trigger keywords.
2026-03-23 12:23:02 +08:00
lizikk
1f9907df4c
ralph-loop[agent-skills-dev]: checkpoint before iteration
2026-03-23 11:24:19 +08:00
lizikk
29bcbb89db
feat: add claude-md-guardian, thinking commands, and security fixes
...
- Add claude-md-guardian agent with SessionStart hook for auto CLAUDE.md maintenance
- Add 6 thinking commands: /my-world, /emerge, /challenge, /connect, /today, /close
- Add my-world skill for one-shot vault context loading
- Fix command injection vulnerability in init-vault.sh (use env vars)
- Add error handling and logging to vault-sync.sh
- Update write-article skill with complete frontmatter fields
- Upgrade CLAUDE.md to v1.3.0 with cycle time tracking and exit criteria
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
2026-03-02 16:58:43 +08:00
lizikk
c3209dfa68
init: content-forge project with write-article skill, scripts, and CLAUDE.md
2026-03-02 11:20:38 +08:00
