Community skills for Claude Code agents. Built from real production use — extracted from a working AI agent harness, sanitized, and published here.
| Skill | What It Does | Triggers |
|---|---|---|
| psx-invest | Pakistan Stock Exchange live research — KMI-30 Shariah picks with buy/avoid reasoning | PSX, KMI-30, Pakistan stocks, halal investing |
| humor-writing | 5 research-backed humor techniques (BVT, Rule of Three, self-deprecation, authentic voice) | "make this funny", funny LinkedIn post, witty tweet |
| job-agent | Autonomous freelance job lifecycle: scan → score → auto-apply ≥75 → Notion tracking | "apply 1/2/3", "what jobs came in", "write a proposal" |
| orchestrator-harness | SQLite tick-lock orchestration + deterministic event IDs + plan-first approval gate | AI agent harness, autonomous agent, tick-lock |
| kamil-evolution-agent | Failure-driven self-improvement: failures.jsonl → pattern classification → rule rewriting | "fix your behavior", "keep getting X wrong", self-improvement loop |
| notion-brain-agent | Notion-as-brain: 8 typed DBs, rate-limit-safe polling, session-start context injection | Notion as memory, agent brain, Notion automation |
| content-scheduler | Daily content pipeline: Notion queue → research → LLM generation → publish → log | automate posting, daily content, content cron |
| django-lms-bug-agent | Plan-first Django bug lifecycle: hard anti-patterns, E2E gate, quality thresholds | Django bug fix, LMS feature, plan first |
| rocky-voice | Make Claude speak like Rocky from Project Hail Mary — dropped articles, "is" constructions, pure engineering enthusiasm | "talk like Rocky", sci-fi voice, LinkedIn hooks with personality |
| asset-studio | Generate game sprites, icons, Lottie animations, and GIFs entirely as code — parameterized SVG generators, eased motion, contact-sheet critique, SVG→Lottie/sprite-sheet export | "draw me X", "make game assets", "create an animated icon/GIF", "convert SVG to Lottie/sprite sheet" |
In Claude Code, skills in .claude/skills/ are auto-loaded. Copy the whole skill directory into your project (some skills, like asset-studio, ship helper scripts and reference docs alongside SKILL.md — not just the single file):
mkdir -p /your-project/.claude/skills/<skill-name>
cp -r <skill-name>/. /your-project/.claude/skills/<skill-name>/Or reference directly via SkillsMP:
skillsmp.com/creators/oyekamal/agent-skills/<skill-name>
- Extracted from production — not theoretical. Every pattern was tested on real systems.
- One file where possible — no framework, no package.json. A few skills (e.g.
asset-studio) ship small standalone helper scripts because the task genuinely needs runnable code, but never a dependency beyond common pip/npm packages. - Opinionated — each skill makes a choice and explains why. If you want options, read the docs.
- Anti-patterns included — what NOT to do is as important as what to do.
@oyekamal — Senior Django/React developer + AI agent builder