Utility Skills

Complete reference for all 8 utility skills — extend your agent beyond trading into payments, GitHub, hardware, and more

The 8 utility skills extend Gemach agents beyond DeFi trading into a broader range of autonomous capabilities. They are designed to complement GDEX trading skills in compound workflows — for example, summarizing a trade report and pushing it to a GitHub repository, or monitoring hardware health and suspending the agent (entering hibernation mode) if the system overheats.

Skill Catalog

skill-creator

Purpose: A meta-skill that helps agents and users design, scaffold, and write new SKILL.md files.

Key capabilities:

• Accept a natural-language description of a desired capability

• Generate a complete, correctly structured SKILL.md scaffold

• Suggest appropriate tools, parameters, and error conditions

• Create matching helper script stubs (Node.js, Python, or shell)

• Validate the output against the SKILL.md format specification

Why it matters: This skill is the entry point for growing the skills ecosystem. Instead of writing a new skill from scratch, you describe what you want and the agent drafts the structure for you. It bootstraps the community contribution loop.

Example usage:

You: Create a skill that monitors a Telegram bot for incoming messages
     and forwards them to a webhook URL.

Agent: I've scaffolded workspace/skills/telegram-webhook/SKILL.md with
       a listen_for_messages tool and helpers/poll-telegram.js. Review
       the file and let me know what to adjust.

Example prompts:

"Create a new skill for fetching DEX prices from Birdeye."
"Scaffold a skill that lets me interact with the Airtable API."
"Help me write a skill for monitoring my Solana wallet for incoming transfers."

---

github

Purpose: Interact with GitHub repositories as part of agent-automated workflows.

Key capabilities:

• Create, read, update, and close issues

• Post comments on issues and pull requests

• Push file changes to a repository branch

• Create and merge pull requests

• Read file contents from any accessible repository

• List repository branches, commits, and open PRs

Common use cases:

Workflow	Description
Trade log publishing	Auto-commit daily trade reports to a private repo
Bug reporting	Open GitHub issues when SDK errors are detected
Strategy versioning	Commit updated strategy parameters to version control
Docs updating	Push skill documentation updates automatically

Example prompts:

"Open a GitHub issue in my-org/my-repo with the title 'SDK rate limit exceeded'."
"Commit today's trade log to the reports/ directory in my trading-logs repo."
"Show me all open PRs in GemachDAO/Gclaw."
"Post a comment on PR #42 in my-repo saying the tests passed."

Authentication: Requires a GitHub personal access token (PAT) or GitHub App credentials configured in the agent's environment.

---

summarize

Purpose: Condense any text — documents, trade logs, chat histories, market reports — into structured, actionable summaries.

Key capabilities:

• Summarize long documents to a configurable length (brief, detailed, or bullet points)

• Extract key insights, decisions, and action items from conversation logs

• Generate trade recap reports from raw transaction histories

• Condense market analysis reports into executive summaries

• Multi-document summarization with cross-source synthesis

Output formats:

Format	Use Case
brief	2–3 sentence summary for quick status checks
bullets	Bulleted list of key points — best for action items
detailed	Structured multi-paragraph summary for full context
report	Formatted report with sections and headings

Example prompts:

"Summarize my last 7 days of trade activity into a weekly report."
"Give me a 3-bullet summary of this market analysis."
"Condense this 5000-word whitepaper into the key points."
"Create a summary of today's agent activity log."

---

hardware

Purpose: Monitor and respond to hardware health metrics on the host system — especially useful for agents running on low-resource boards.

Key capabilities:

• Monitor CPU usage, memory usage, disk space, and temperature

• Set alert thresholds that trigger agent actions (slow down inference, hibernate, send notification)

• Query current system metrics on demand

• Log hardware health over time for trend analysis

• Trigger graceful hibernation when resources are critically low

Threshold actions:

Condition	Default Threshold	Default Action
CPU > 90% for 60s	90%	Pause non-critical tasks
Memory > 85%	85%	Free caches, reduce context window
Temperature > 80°C	80°C	Throttle inference; alert
Disk < 5% free	5%	Alert; do not write new files

Example prompts:

"What's the current CPU and memory usage?"
"Alert me if the temperature on my Raspberry Pi exceeds 75°C."
"Check if I have enough free disk space to store today's trade logs."
"Put the agent in low-resource mode — I'm running other workloads."

Typical deployment: Hardware monitoring is most valuable on agents running on single-board computers (Raspberry Pi, Orange Pi, Radxa) where thermal and memory limits are real constraints.

---

weather

Purpose: Fetch current weather conditions and forecasts for use in multi-agent workflows and context-aware decision-making.

Key capabilities:

• Current conditions: temperature, humidity, wind, UV index

• Hourly and 7-day forecasts

• Weather alerts and severe conditions

• Location lookup by city name, coordinates, or IP geolocation

Use in compound workflows:

Weather data becomes useful when combined with other skills. For example, an agent managing agricultural equipment might check weather before scheduling irrigation; or a DeFi event agent might note weather conditions when logging real-world context for a trade journal.

Example prompts:

"What's the weather in San Francisco right now?"
"Will it rain in London tomorrow afternoon?"
"Get the 7-day forecast for 37.7749,-122.4194."
"Are there any weather alerts for New York City?"

---

tmux

Purpose: Create, attach to, and manage tmux sessions — enabling agents to run multiple parallel processes in a structured terminal environment.

Key capabilities:

• Create new named tmux sessions and windows

• Split panes horizontally and vertically

• Send commands to specific panes without attaching

• Capture pane output for reading by the agent

• List all active sessions and their status

• Kill specific sessions or windows

Typical use cases:

Use Case	Description
Multi-agent orchestration	Run multiple Gclaw agents side by side in named sessions
Log monitoring	Keep a log-tail window alongside the main agent
Background processes	Keep helper processes alive across agent restarts
Development workflows	Run test suite, watcher, and REPL in parallel panes

Example prompts:

"Start a new tmux session called 'trading-agent'."
"Open a second window in my current tmux session for monitoring logs."
"Send 'gclaw monitor' to the second pane."
"Show me the output from the log-monitor pane."
"Kill the old development session."

---

tempo-payment

Purpose: Handle Tempo payment protocol flows — enabling agents to send, receive, and verify Tempo-based payments as part of automated transaction workflows. Tempo is a programmable payment protocol built for autonomous agents, allowing machines to transact with each other using on-chain payment channels.

Key capabilities:

• Initiate Tempo payment requests with configurable amounts and recipients

• Monitor payment status and confirm receipt

• Integrate Tempo payments into multi-step agent workflows

• Handle payment callbacks and webhooks

• Record payment history and receipts

Example prompts:

"Send a Tempo payment of 10 USDC to wallet 0xRecipient..."
"Check if payment ID tempo_xyz123 has been confirmed."
"Set up a recurring Tempo payment of 5 USDC per day to my service wallet."

---

x402

Purpose: Process x402 micropayment protocol flows — letting agents autonomously pay for external APIs and services without human intervention.

Key capabilities:

• Detect x402 payment challenges from HTTP 402 responses

• Automatically construct and sign payment proofs

• Pay for API access with configurable per-request budget limits

• Track cumulative spend across all x402-protected endpoints

• Reject payment requests that exceed configured spend limits

What x402 enables: The x402 protocol extends HTTP with native micropayment support. When an agent requests a resource and receives a 402 Payment Required response, the x402 skill handles the payment challenge automatically — the agent pays and retries without human involvement. This enables truly autonomous access to paid data feeds, AI inference endpoints, and other metered services.

Example prompts:

"Fetch the premium market data report — handle the payment automatically."
"What's my total x402 spend this session?"
"Set a budget limit of 1 USDC per hour for all x402 payments."
"Use x402 to access the Chainlink premium data feed."

---

Combining Utility and Trading Skills

Utility skills are designed to compose naturally with GDEX trading skills. Some powerful combinations:

Combination	Workflow
gdex-token-discovery + summarize	Discover trending tokens, summarize findings into a briefing
gdex-portfolio + github	Track portfolio daily and commit reports to a private repo
gdex-spot-trading + x402	Buy tokens using paid data feeds for signal generation
hardware + gdex-trading	Monitor hardware health and pause trading if system overheats
skill-creator + any skill	Use the agent to extend itself with new capabilities
tmux + any skill	Run multiple trading strategies in parallel sessions

---

See also:

• 📈 GDEX Trading Skills — the 14 trading skills reference

• ✏️ Creating Custom Skills — add your own utility skills

• 🧩 Agent Skills Overview — full catalog and loading priority