Model Intel

Live LLM model intelligence and pricing from OpenRouter

Security Scan

The skill generally matches its stated purpose (querying OpenRouter for live model data) but has inconsistent metadata and minor scope creep: its README says an API key is required yet the registry lists none, and the script will read your agent workspace file (~/.openclaw/workspace/.env) for secrets — behavior that isn't clearly justified.

Purpose & Capability

Name/description match the implementation: the script calls openrouter.ai to list/search/compare models and pricing. However, SKILL.md declares an OPENROUTER_API_KEY requirement while the registry metadata lists no required env vars; the script reads for that key but never actually uses it when calling the public models endpoint, which is an inconsistency.

Instruction Scope

The runtime instructions tell users to set OPENROUTER_API_KEY and run the included script — expected. The script, though, attempts to read ~/.openclaw/workspace/.env to extract OPENROUTER_API_KEY if not present in the environment. Accessing an agent workspace file is unexpected for a read-only model-listing tool and constitutes scope creep; while the script only looks for a specific line, reading that file could expose other workspace secrets if the code were modified.

Install Mechanism

No install spec is provided (instruction-only plus a single Python script). That is low-risk from an installation-perspective; required dependency is only the 'requests' Python package which is reasonable.

Credentials

SKILL.md requires OPENROUTER_API_KEY, but the registry metadata declares no required env vars or primary credential — a mismatch. The script's fallback to read the workspace .env file to obtain the key increases the privilege surface (it accesses a local config path), even though the key is not actually used in the observed network calls.

Persistence & Privilege

The skill does not request permanent presence (always:false), does not modify other skills or system settings, and does not persist credentials beyond reading the workspace file — no elevated persistence or privilege is requested.

What to consider before installing

This skill largely does what it says (fetches model listings from openrouter.ai), but there are a few red flags you should consider before installing: (1) SKILL.md says you need OPENROUTER_API_KEY but the registry metadata doesn't declare that — ask the author to correct the metadata; (2) the script will read ~/.openclaw/workspace/.env if the env var is missing, which can expose secrets kept in your agent workspace — avoid storing unrelated secrets there or run the skill in an isolated environment; (3) the code reads the key but does not appear to use it for the models GET request (likely a bug) — confirm whether authenticated endpoints are needed and how the key is used; (4) if you don't trust the publisher, inspect or run the script in a sandbox/container and verify network traffic only goes to openrouter.ai. If the author can justify the workspace file read and fix the metadata (or remove the fallback), the inconsistencies would be resolved.

Live LLM model intelligence from OpenRouter. Compare pricing, search models, find the best model for any task (code, reasoning, creative, fast, cheap, vision, long-context). Real-time data, not stale training knowledge.

Usage

# List top models by provider
python3 scripts/model_intel.py list

# Search by name
python3 scripts/model_intel.py search "claude"

# Side-by-side comparison
python3 scripts/model_intel.py compare "claude-opus" "gpt-4o"

# Best model for a use case
python3 scripts/model_intel.py best fast
python3 scripts/model_intel.py best code
python3 scripts/model_intel.py best reasoning
python3 scripts/model_intel.py best cheap
python3 scripts/model_intel.py best vision

# Pricing details
python3 scripts/model_intel.py price "gemini-flash"

Requirements

• OPENROUTER_API_KEY environment variable
• Python 3.10+
• requests package

Credits

Built by AgxntSix — AI ops agent by M. Abidi 🌐 agxntsix.ai | Part of the AgxntSix Skill Suite for OpenClaw agents