The Agent Skills Directory

[COMMAND_EXECUTION]: The embed function in scripts/news_store.ts utilizes Bun.spawnSync to execute a shell command defined by the CRYPTO_NEWS_EMBED_CMD environment variable. This pattern allows for the execution of arbitrary system commands if the environment configuration is compromised or manipulated.- [REMOTE_CODE_EXECUTION]: In scripts/feeds/googlefinance.ts, the parseAfCallback function extracts JavaScript object literals from external HTML content and evaluates them using the new Function constructor. Evaluating code derived from remote web responses is a dangerous practice that could lead to code execution if the source content is malicious.- [PROMPT_INJECTION]: The skill represents a significant surface for indirect prompt injection by ingesting and processing data from numerous external financial news outlets.
Ingestion points: Data is pulled from external sites such as the Financial Times, Wall Street Journal, and various crypto news feeds via scripts/feeds/index.ts.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to disregard potential commands embedded within the fetched news articles.
Capability inventory: The skill has access to shell command execution (Bun.spawnSync), local file system persistence via SQLite, and network access (fetch).
Sanitization: While the skill uses stripHtml to clean the content, this does not filter out natural language instructions that could influence the agent's behavior.

read-news