email-dns-health

Pass

Audited by Gen Agent Trust Hub on Mar 3, 2026

Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it processes and presents data from external, potentially attacker-controlled DNS TXT records (SPF, DKIM, DMARC).\n
  • Ingestion points: The dig_txt function in scripts/email-dns-health.sh fetches arbitrary text from remote DNS TXT records.\n
  • Boundary markers: While output is structured as JSON, the raw content of DNS records is passed to the agent without delimiters or instructions to ignore embedded commands.\n
  • Capability inventory: The skill allows the agent to read local environment files and provides instructions for performing automated DNS updates via API.\n
  • Sanitization: The script performs basic character filtering using tr and sed but does not sanitize the records for malicious natural language instructions that could influence agent behavior.\n- [DATA_EXFILTRATION]: The skill documentation (SKILL.md) directs the agent to access sensitive credentials (Cloudflare API token) stored at ~/.claude/email-dns-health/.env. While this is a skill-specific configuration file used for the legitimate 'fix' functionality, accessing environment files containing secrets is a sensitive operation that increases the impact of potential injection attacks.\n- [COMMAND_EXECUTION]: The helper script scripts/email-dns-health.sh executes system commands dig and jq using domain names provided by the user. The domain arguments are correctly double-quoted in the shell script, which prevents direct command injection via the shell.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 3, 2026, 09:28 PM