The Agent Skills Directory

[REMOTE_CODE_EXECUTION] (HIGH): The script scripts/crawl.sh executes npx -y mcp-remote, which downloads and runs external code from the npm registry at runtime without integrity verification.
[EXTERNAL_DOWNLOADS] (HIGH): The skill's authentication process depends on downloading the mcp-remote package and connecting to https://mcp.tavily.com/mcp at runtime.
[DATA_EXFILTRATION] (MEDIUM): The script reads sensitive authentication tokens from the ~/.mcp-auth/ directory. Accessing hidden credential caches in the user's home folder is a high-risk operation, though the severity is reduced to MEDIUM as it is required for the skill's stated OAuth functionality.
[PROMPT_INJECTION] (LOW): The skill is susceptible to indirect prompt injection as it retrieves untrusted data from the web and writes it directly to local markdown files without sanitization.
Ingestion points: External website content fetched via curl in scripts/crawl.sh.
Boundary markers: None identified.
Capability inventory: Local file-write operations in scripts/crawl.sh.
Sanitization: No validation or sanitization is performed on the content of the crawled pages.
[COMMAND_EXECUTION] (LOW): The script uses standard shell utilities including curl, jq, sed, and mkdir to perform its tasks.

crawl