article-fetch
Warn
Audited by Gen Agent Trust Hub on Nov 12, 2026
Risk Level: MEDIUMINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill's primary function is to fetch content from arbitrary external URLs (X/Twitter, WeChat, web pages) and local files. The instructions explicitly direct the agent to "read through article.md" once the content is fetched. This creates a surface where malicious instructions embedded in a web article or social media post could influence the agent's behavior.
- Ingestion points:
fetch.mjsdownloads content from user-provided URLs or reads local files intoarticle.md. - Boundary markers: The skill does not define specific delimiters or instructions to ignore embedded prompts within the fetched content.
- Capability inventory: The skill has access to the
Bashtool to run node scripts, write files, and executeffmpeg. - Sanitization: The documentation does not specify sanitization procedures for the content retrieved from external sources.
- [DYNAMIC_EXECUTION]: The wrapper scripts
scripts/doctor.mjsandscripts/fetch.mjsuse dynamic imports to load their core logic. The import path is determined by theARTICLE2MEDIA_ENGINEenvironment variable or a relative path../../article2media/engine/scripts/. - Evidence:
await import(new URL('doctor.mjs', engine))andawait import(new URL('fetch.mjs', engine))are used to execute code from a computed path. - [COMMAND_EXECUTION]: The skill uses the
Bashtool to execute Node.js scripts. User-supplied arguments, including URLs and directory paths, are passed directly to these scripts. - Evidence:
node ${CLAUDE_SKILL_DIR}/scripts/fetch.mjs "<url 或文件>"allows user input to flow into a shell command. - [EXTERNAL_DOWNLOADS]: The skill fetches content from external sources as part of its conversion process. It mentions using
r.jina.aianddefuddle.mdas third-party services to process web pages.
Audit Metadata