skill-web-importer

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt instructs fetching raw SKILL.md and its reporting format explicitly includes "Pattern: """ for detected secrets/hardcoded tokens, which requires the LLM to reproduce secret values verbatim in its output.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 1.00). The skill explicitly fetches raw SKILL.md (and referenced ref/ files) from external URLs via WebFetch (e.g., GitHub, ClawHub or any host) and the LLM is required to read and analyze that untrusted, user-sourced content to decide PASS/WARN/FAIL, which exposes the agent to indirect prompt injection risks.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly performs a runtime WebFetch of the source SKILL.md (e.g. https://raw.githubusercontent.com///main/SKILL.md) and uses that fetched content as required input to the LLM analysis, so a malicious SKILL.md could contain prompt-injection or encoded execution instructions that would directly influence agent prompts or trigger remote code paths.