Case 05437
A fast, offline static risk analysis CLI for AI agent skill files. Detects malicious instructions, steganographic payloads, and dangerous capability chains.
npx -y skills add knownasnaffy/prompthound --skill case_05437Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
- 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Analyze OpenClaw agent configuration and API usage patterns to identify cost-saving opportunities. Diagnose inefficient heartbeat configs, estimate daily/weekly API spend, and generate actionable recommendations to reduce LLM API costs. Essential for any agent operator running 24/7 on a budget. Use when: agent costs are higher than expected, heartbeat config needs review, or monthly API bill surprises you.
SKILL.md
4.2 KB, as published. Nobody here has run it
API Cost Optimizer
Diagnose, estimate, and optimize LLM API costs for OpenClaw agents running 24/7.
What This Skill Does
- Estimates API spend from agent configuration (model, heartbeat interval, task complexity)
- Diagnoses heartbeat waste — the #1 cause of runaway API bills
- Counts and categorizes tools the agent has access to
- Produces a cost report with prioritized recommendations
- Estimates time-to-payback for any recommended changes
The Problem This Solves
Documented cases show agents burning $8–$250 per week in API costs due to misconfigured heartbeats and inefficient task patterns. A properly tuned heartbeat alone can reduce costs by 60–90%.
Environment Variables
| Variable | Required | Default | Description |
|---|---|---|---|
API_COST_MODEL | No | openai | Provider: openai, anthropic, minimax, lmstudio |
API_COST_INTERVAL | No | weekly | Report interval: daily, weekly, monthly |
HEARTBEAT_INTERVAL | No | auto | Override heartbeat interval in seconds (auto-detect if not set) |
MODEL_PRICE_INPUT | No | auto | Price per 1M input tokens (auto-selected by provider) |
MODEL_PRICE_OUTPUT | No | auto | Price per 1M output tokens (auto-selected by provider) |
Provider Default Pricing (per 1M tokens)
| Provider | Model | Input | Output |
|---|---|---|---|
| OpenAI | GPT-4o | $2.50 | $10.00 |
| OpenAI | GPT-4o-mini | $0.15 | $0.60 |
| OpenAI | GPT-4-turbo | $10.00 | $30.00 |
| Anthropic | Claude 3.5 Sonnet | $3.00 | $15.00 |
| Anthropic | Claude 3 Opus | $15.00 | $75.00 |
| MiniMax | MiniMax-M2 | $0.10 | $0.10 |
| MiniMax | MiniMax-M2.1 | $0.20 | $0.40 |
Scripts
analyze.sh — Full Cost Analysis
Runs a complete cost diagnostic on the current OpenClaw configuration.
./scripts/analyze.sh
Output: Detailed markdown report with:
- Estimated API spend (daily / weekly / monthly)
- Heartbeat cost analysis
- Tool count risk factor
- Top 3 cost reduction recommendations
- Estimated savings after optimization
heartbeat_diagnosis.sh — Find Heartbeat Waste
Checks OpenClaw heartbeat configuration and calculates wasted API calls.
./scripts/heartbeat_diagnosis.sh
Output: Heartbeat efficiency score (0–100%), wasted calls per day, estimated annual waste.
estimate.sh — Quick Cost Estimate
One-shot estimate with optional custom parameters.
./scripts/estimate.sh <heartbeat_seconds> <tasks_per_day> <avg_input_tokens> <avg_output_tokens>
Output: Daily, weekly, and monthly cost estimates for the given parameters.
Cost Reduction Recommendations
Priority order:
- Increase heartbeat interval — If heartbeat is under 5 minutes, increase to 10–30 min for idle agents
- Switch to cheaper model — MiniMax M2 at $0.10/1M is 25x cheaper than GPT-4o for most tasks
- Batch tasks — Run 5–10 tasks per session instead of 1 task per heartbeat cycle
- Cache responses — Use memory skills to avoid repeated context
- Disable verbose logging — Reduces token churn in long conversations
Usage Example
# Full analysis with MiniMax pricing
export API_COST_MODEL=minimax
./scripts/analyze.sh
# Quick estimate
./scripts/estimate.sh 300 20 8000 2000
# Heartbeat-only diagnosis
./scripts/heartbeat_diagnosis.sh
Notes
- Estimates are based on configuration analysis, not live API metering
- For actual spend tracking, enable provider cost logs and compare
- MiniMax models offer best price/performance for volume workloads
- Heartbeat waste is typically the single largest cost lever