skill-installer

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

• Third-party content exposure detected (high risk: 0.72). The tool reads outsider-authored free text from other skills’ SKILL.md files listed in marketplace.json (via fs.readFile(skillMdPath, "utf-8") in getMarketplaceSkills()), then returns the extracted description/name as JSON text in the search_skills tool output (content: [{type:"text", text: JSON.stringify(results)}]), which can be ingested into the agent’s LLM context.

MEDIUM W021: Hidden or invisible Unicode characters detected (potential obfuscation or prompt injection).

• Hidden Unicode characters detected (1 type(s) found)