blogwatcher
Warn
Audited by Gen Agent Trust Hub on Sep 1, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill provides instructions to download binary executables directly from a personal GitHub repository (github.com/JulienTant/blogwatcher-cli). While the author matches the skill author, these resources do not originate from a verified trusted organization or official registry.- [REMOTE_CODE_EXECUTION]: The installation guide suggests piping remote downloads directly into a system binary path (/usr/local/bin) using 'curl | tar'. This method executes unverified third-party binaries on the host system without integrity checks.- [COMMAND_EXECUTION]: The skill relies on the execution of the 'blogwatcher-cli' tool to perform its primary functions, including scanning blogs, managing articles, and importing OPML files.- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external RSS/Atom feeds and scraped HTML content. This content is displayed to the agent via the 'articles' command, potentially allowing an attacker to influence the agent's behavior through malicious instructions embedded in blog content.
- Ingestion points: RSS/Atom feeds and blog URLs processed by 'blogwatcher-cli'.
- Boundary markers: No explicit delimiters or safety instructions are provided to the agent to distinguish between tool output and untrusted content.
- Capability inventory: The skill uses 'blogwatcher-cli' to interact with a local SQLite database and network resources.
- Sanitization: No evidence of sanitization or escaping of external article content is present in the skill instructions.
Audit Metadata