skill-installer
Pass
Audited by Gen Agent Trust Hub on Apr 6, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill is susceptible to Indirect Prompt Injection. The
search_skillsandgetMarketplaceSkillsfunctions readSKILL.mdfiles from paths defined in an externalmarketplace.jsonfile. It extracts the name and description metadata from these files and returns them to the agent context. - Ingestion points:
index.tsreads content fromSKILL.mdfiles located at paths specified in../../../.claude-plugin/marketplace.json. - Boundary markers: Absent. The parsed metadata is returned as a raw JSON string to the agent.
- Capability inventory: The skill has permissions to read and write to the local filesystem (
fs.readFile,fs.writeFile,fs.access). - Sanitization: Absent. The code naively extracts strings from lines starting with '#' or the first non-header line without any validation or escaping of the content.
Audit Metadata