The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process external DNS records and HTTP response data, which constitutes an attack surface for indirect prompt injection.
Ingestion points: The agent ingests data from DNS TXT records, HTTP headers, and response bodies during audits (e.g., in references/mail-authentication.md and references/web-routing-and-tls.md).
Boundary markers: The skill instructions in SKILL.md (Safety section) explicitly command the agent to treat all remote content as untrusted evidence rather than instructions and forbid executing commands or following links based on this data.
Capability inventory: The agent has capabilities to write to a local user cache, execute system shell commands (curl, nslookup, node), and perform authenticated mutations via the Cloudflare API.
Sanitization: A dedicated Node.js utility (scripts/cache-store.mjs) implements strict sanitization, using regex patterns to detect and block the caching of JWTs, API keys, private keys, and session cookies.
[COMMAND_EXECUTION]: The skill requires the agent to execute shell commands to perform its auditing functions.
Evidence: Template recipes for Resolve-DnsName, curl.exe, and nslookup.exe are documented in references/command-recipes.md.
Mitigation: The skill enforces atomic validation of all hostname inputs before processing and requires passing hostnames as command arguments rather than interpolated shell strings to prevent shell injection attacks.

dns-doctor