The Agent Skills Directory

[External Downloads] (LOW): The UpdateSources.ts tool fetches data from a third-party GitHub repository (jacobdjwilson/awesome-annual-security-reports) not listed in the trusted sources. While it only parses the content for URLs, it establishes a dependency on an external, untrusted source for the skill's core data.\n- [Data Exposure & Exfiltration] (LOW): The FetchReport.ts tool includes a --url flag that allows the agent to fetch content from any provided URL. This creates a surface for Server-Side Request Forgery (SSRF) if the agent is used to probe internal network resources or fetch sensitive metadata from cloud environments.\n- [Prompt Injection] (LOW): The skill is susceptible to Indirect Prompt Injection (Category 8) because it ingests untrusted data from security reports and external README files.\n
Ingestion points: Tools/FetchReport.ts (external report content) and Tools/UpdateSources.ts (remote GitHub README).\n
Boundary markers: Absent. The summary files are generated by appending raw extracted text directly into markdown files, providing no delimiters to help the LLM distinguish between instructions and data.\n
Capability inventory: The skill can execute commands via bun, read and write local files in the Reports/ directory, and perform network requests via fetch.\n
Sanitization: The skill performs basic HTML tag stripping using regex, but it does not sanitize or filter for embedded LLM instructions within the fetched content.

AnnualReports