The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill exhibits a significant attack surface for indirect prompt injection by processing untrusted data from multiple external sources.
Ingestion points: distribb_research.py performs web scraping of arbitrary URLs; news_topics.py fetches Google News RSS feeds; commands/gbp.md triages live Google reviews; and commands/review-video.md scrapes competitor reviews from platforms like Trustpilot, G2, and Reddit.
Boundary markers: The instructions lack clear delimiters or specific 'ignore embedded instructions' warnings when the ingested data is passed to the LLM for analysis or writing.
Capability inventory: The agent possesses powerful capabilities including network operations via curl, CMS article publishing (POST /articles), social media posting (POST /social/publish), and public business review replies (POST /gbp/reviews/reply).
Sanitization: While distribb_research.py removes HTML tags using BeautifulSoup, it does not filter the text content for potential injection payloads.
[DATA_EXPOSURE]: The skill exposes administrative API endpoints that increase the discoverable attack surface of the platform.
distribb_cli.py contains command-line wrappers for internal administrative endpoints such as /api/v1/admin/articles and /api/v1/admin/landers.
[COMMAND_EXECUTION]: One of the provided utility scripts explicitly disables security protections during network requests.
distribb_research.py uses requests.get(..., verify=False) when scraping data, which makes the collection process vulnerable to Man-in-the-Middle (MitM) attacks.
[EXTERNAL_DOWNLOADS]: The skill relies on external package installations and cross-skill dependencies.
The installation process uses npx skills add Bomx/distribb-skill.
The skill further requires the installation of Bomx/super-video-maker-skill for its video explainer features.

distribb