Production-grade Agent Skills for OpenClaw and the ClawHub marketplace.
Built by CacheForge — making AI agents affordable, powerful, and production-ready.
Real CLI integrations. Safe by default. Discord v2 ready. 30M+ tokens battle-tested.
| Skill | Description | Version | Install |
|---|---|---|---|
| vibe-check | Code quality auditor — catches "vibe coding" sins | 0.1.1 | clawhub install vibe-check |
| dep-audit | Dependency vulnerability scanner (npm, pip, Cargo, Go, Ruby) | 0.1.3 | clawhub install dep-audit |
| rug-checker | Solana token rug-pull risk analysis (10-point scoring) | 0.1.3 | clawhub install rug-checker |
| Skill | Description | Version | Install |
|---|---|---|---|
| prom-query | Prometheus metrics query & alert interpreter | 1.0.1 | clawhub install prom-query |
| kube-medic | Kubernetes cluster triage & pod autopsy | 1.0.1 | clawhub install kube-medic |
| log-dive | Unified log search (Loki / Elasticsearch / CloudWatch) | 0.1.1 | clawhub install log-dive |
| tf-plan-review | Terraform plan risk assessment & blast radius analysis | 0.1.1 | clawhub install tf-plan-review |
| Skill | Description | Version | Install |
|---|---|---|---|
| pager-triage | PagerDuty incident triage & resolution workflows | 0.1.1 | clawhub install pager-triage |
| Skill | Description | Version | Install |
|---|---|---|---|
| feed-diet | Information diet auditor (HN, RSS/OPML analysis) | 0.1.1 | clawhub install feed-diet |
| meeting-autopilot | Meeting transcript → action items, decisions & follow-ups | 0.1.1 | clawhub install meeting-autopilot |
| Skill | Description | ETA |
|---|---|---|
| sentry-scout | Sentry error triage & stacktrace analysis | Mon–Tue |
| test-pilot | Playwright test runner & failure analyzer | Wed |
| stripe-dash | Stripe revenue intelligence (read-only) | Thu–Fri |
# Install from ClawHub
clawhub install vibe-check
# Or clone and install any skill locally
git clone https://github.com/cacheforge-ai/cacheforge-skills.git
cp -r cacheforge-skills/skills/vibe-check ~/.openclaw/skills/🔧 Real CLI integrations — wraps actual tools (promtool, kubectl, terraform, pip-audit, cargo-audit, npm audit), not prompt-only LLM wrappers that hallucinate data.
🛡️ Safe by default — read-only first. Every destructive action is gated by explicit user confirmation. No YOLO kubectl delete on your production cluster.
💬 Discord v2 ready — compact first response with interactive follow-up actions. Designed for OpenClaw v2026.2.14+ delivery modes across Discord, Telegram, and webchat.
🏗️ Security hardened — no eval, no bash -c with variable interpolation, all JSON via jq --arg, URL scheme validation, set -euo pipefail everywhere. Every script passes bash -n syntax checks.
📊 Screenshot-worthy output — scored reports, risk matrices, ASCII charts, and emoji-rich Markdown. Built to be shared.
⚡ Battle-tested — this entire skill suite was built, reviewed, and hardened in a single day using a multi-agent pipeline running through CacheForge. 30M+ tokens. 455 requests. 10 production skills.
CacheForge is the only publisher on ClawHub with a complete observability suite:
┌─────────────┐ ┌──────────┐ ┌───────────────┐ ┌──────────────┐
│ prom-query │ │ log-dive │ │ pager-triage │ │ sentry-scout │
│ (metrics) │ │ (logs) │ │ (incidents) │ │ (errors) │
└──────┬───────┘ └────┬─────┘ └──────┬────────┘ └──────┬───────┘
│ │ │ │
└───────────────┴───────────────┴───────────────────┘
Your AI-powered NOC
Query Prometheus metrics → correlate with logs → triage the PagerDuty alert → trace the Sentry error. All through your OpenClaw agent.
All CacheForge skills include a Discord v2 delivery contract for OpenClaw v2026.2.14+:
- 📦 Compact first response — summary with key findings, under ~1200 chars
- 🎛️ Component-style follow-ups — interactive quick actions when Discord components are available
- 📋 Graceful fallback — numbered action list when components aren't available
- 📄 Chunked detail — long reports sent in ≤15 line chunks for readability
Works great on Telegram (with inline buttons), webchat, and all other OpenClaw surfaces.
Every CacheForge skill follows the same contract:
SKILL.md → Agent-readable skill definition (frontmatter + instructions)
README.md → Human-readable documentation
CHANGELOG.md → Version history
SECURITY.md → Security model and threat analysis
TESTING.md → Test procedures and validation
scripts/ → Bash CLI wrappers (set -euo pipefail, no eval)
Skills are stateless — no databases, no background processes, no daemons. They wrap real CLI tools and APIs, adding agent intelligence on top.
This entire skill suite was built using a multi-agent pipeline running through CacheForge:
| Metric | Value |
|---|---|
| Total tokens | 30.6M |
| Requests | 455 |
| Model | Claude Opus 4.6 |
| Direct API cost | ~$413 |
| CacheForge cost | ~$143 |
| Savings | $270 (65%) |
CacheForge makes AI agents affordable at scale. Learn more →
We're not accepting external contributions yet. Open an issue to suggest a skill.
Built with 🔥 by CacheForge
10 skills. 30M tokens. One day. One platform.