skills/janyork/llm-wiki-cli/using-lwc/Gen Agent Trust Hub

using-lwc

Pass

Audited by Gen Agent Trust Hub on Oct 3, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill automates the installation of its core binary (lwc) via scripts/install-lwc.sh, which downloads the executable from the author's GitHub repository (JanYork/llm-wiki-cli). The process includes a mandatory SHA-256 checksum verification step to ensure the integrity of the downloaded file before it is executed.\n- [EXTERNAL_DOWNLOADS]: The skill references and provides commands to install optional document conversion engines, specifically Microsoft's markitdown and Firecrawl's @firecrawl/anydoc. These are well-known, legitimate packages intended for processing PDF, Office, and other non-markdown formats.\n- [INDIRECT_PROMPT_INJECTION]: Because the skill is designed to ingest and summarize external documents and Wiki pages, it presents an attack surface for indirect prompt injection. The skill provides clear defensive instructions in SKILL.md and references/active-memory.md, directing the agent to treat all ingested content as untrusted reference data that cannot override system or developer policies.\n- [COMMAND_EXECUTION]: The skill performs project-level operations by executing the lwc CLI tool. It uses scripts/bootstrap.sh to identify valid project roots and verify command availability, implementing checks to ensure operations remain within host-authorized directories.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 3, 2026, 09:11 PM