Trust Decay Monitor

Helps track how AI skill verification results decay over time. A "verified" badge from 18 months ago may be meaningless today — dependencies updated, new att...

This skill is internally coherent and lightweight: it only needs curl and python3 and is instruction-only. Before installing, consider: (1) the skill will fetch external resources (CVE feeds, marketplaces, endpoints) — if you don't want the agent making network calls, restrict network egress or run checks in a controlled environment; (2) the skill does not request credentials, but if you later ask it to check private marketplace profiles you will need to provide appropriate read-only API keys — prefer scoped, short-lived credentials; (3) trust scores are heuristic: use them to prioritize re-audits, not as definitive evidence of compromise; (4) because the skill can be invoked autonomously, be aware it may perform queries automatically if allowed — limit autonomous execution if you want to control when network checks occur.

Type: OpenClaw Skill Name: trust-decay-monitor Version: 1.0.0 The skill's stated purpose is to monitor the decay of trust in AI skill verifications, which is a legitimate security-focused function. The `SKILL.md` file clearly describes this purpose, its inputs, outputs, and methodology without any evidence of prompt injection attempts, malicious instructions, or intent to subvert the agent. The required binaries (`curl`, `python3`) are general-purpose and plausibly needed for a security monitoring tool that would interact with various data sources, and their mere presence does not indicate malicious intent in this context. No other files suggest any harmful behavior.