The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Bash tool in Phase 4 to perform PDF conversion by invoking Google Chrome or npx. These commands incorporate variables such as {token_name}, which are derived from user input or external search results. The lack of explicit sanitization for these variables during shell command construction presents a risk of command injection.
[EXTERNAL_DOWNLOADS]: Phase 4 uses npx -y md-to-pdf to fetch and execute a utility from the npm registry during the runtime execution of the skill.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface by ingesting data from external URLs (via WebFetch) and search results (via WebSearch) and interpolating this untrusted content into sub-agent prompts without sanitization or boundary markers.
Ingestion points: WebFetch is used in Agent 2 (news) and Agent 5 (project websites); WebSearch is used in Phase 0 and all five sub-agents.
Boundary markers: No specific delimiters or instructions to ignore embedded commands are present when passing web content to sub-agents.
Capability inventory: The main agent possesses Bash (shell execution) and Write (file system access) capabilities.
Sanitization: No validation or escaping of external content is performed before processing.

analyze-crypto