The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches content from well-known and reputable technical domains such as Hacker News (news.ycombinator.com), HuggingFace (huggingface.co), and Paul Graham's website (paulgraham.com). These external references are consistent with the skill's core purpose of technical news curation and do not involve suspicious or high-risk domains.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface (Category 8) because it ingests and processes untrusted data from external websites.
Ingestion points: Data is ingested from multiple external URLs using the WebFetch tool and mcp__chrome-devtools__ browser snapshot tools as defined in SKILL.md and sources.json.
Boundary markers: There are no explicit instructions, delimiters, or warnings defined in the orchestrator prompts to isolate the fetched content or to instruct the model to ignore adversarial instructions found within the scraped HTML or text.
Capability inventory: The skill utilizes Write, Bash (limited to mkdir, date, and ls), and mcp__chrome-devtools__ tools across its execution phases, providing a surface for actions based on interpreted content.
Sanitization: While the skill performs extraction and quality scoring, it does not specify procedures for sanitizing, escaping, or filtering instruction-like content from the scraped text before it is processed by the language model for report generation.

daily-news-report