Skillv1.0.0

ClawScan security

Macro Regime Detector · ClawHub's context-aware review of the artifact, metadata, and declared behavior.

Scanner verdict

SuspiciousFeb 21, 2026, 7:38 PM

Verdict: suspicious
Confidence: medium
Model: gpt-5-mini
Summary: The skill's stated purpose (multi-source macro regime detection) is plausible, but the runtime instructions are vague, contain internal inconsistencies, and leave the agent broad discretion to fetch and combine external data without specifying sources, weights, or required credentials.
Guidance: This skill is plausible but underspecified. Before installing or using it: 1) Ask the author to list exact data sources/endpoints and whether API keys or paid access are required (e.g., which news API, on-chain provider, Reddit method). 2) Request a clear algorithm: how signals are normalized, how weights are combined, and tiebreaker rules (explain why F&G=8 produced Regime B in the example). 3) If the agent will scrape websites (Reddit, news), consider legal/usage limits and potential IP/rate concerns. 4) Be cautious about using outputs for automated trading — validate with historical backtests and guardrails. If the author cannot clarify data sources, credentials needed, and the decision logic, treat the skill as unreliable and avoid exposing sensitive credentials or letting it act autonomously on financial systems.

Review Dimensions

Purpose & Capability: noteThe declared purpose (detect macro regimes using Fear & Greed, DXY, VIX, gold/BTC, Reddit sentiment, news, on-chain flows) matches what the SKILL.md describes. No binaries, installs, or credentials are requested, which is not inherently wrong for an instruction-only skill; however several listed signal sources (news NLP, on-chain exchange flows, some premium analytics) commonly require API access or paid data feeds — the skill does not document how those will be obtained. That omission makes it unclear whether the skill can deliver the promised capability without additional secrets or services.
Instruction Scope: concernThe SKILL.md gives a high-level list of signals and returns (regime + confidence + 3 supporting signals + recommendation) but does not define how signals are fetched, normalized, or weighted. It contains an internal inconsistency: the example session lists Regime B (Neutral/Cooling) while showing Fear & Greed = 8 ("Extreme Fear"), which the regime table otherwise maps toward risk-off — this suggests underspecified override rules or contradictory logic. The instructions also give the agent broad discretion to query Reddit, news, and on-chain data without constraints or explicit endpoints, which could lead to unpredictable or excessive data collection behaviors.
Install Mechanism: okInstruction-only skill with no install spec and no bundled code files. This minimizes disk-write risk and there is no external installer to review.
Credentials: noteThe skill requests no environment variables or credentials, which is proportionate on the surface. In practice, some signal sources referenced (news APIs, premium on-chain analytics, robust Reddit access) often require API keys or paid subscriptions; the absence of declared credentials means either the skill expects unauthenticated scraping (fragile/legal/ethical concerns) or the author omitted required auth details. Confirming the intended data sources and credential needs is advised.
Persistence & Privilege: okalways is false and the skill is user-invocable; autonomous model invocation is allowed (default) but not combined with elevated privileges. The skill does not request system-wide configuration changes or persistent installation artifacts.