sec-fetch

Warn

Audited by Gen Agent Trust Hub on Sep 17, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes a local bash script using arguments that are directly interpolated into a command string.
  • Evidence: bash ${CLAUDE_PLUGIN_ROOT}/skills/sec-fetch/scripts/sec-fetch.sh "<url>" "<sec.user>" "<sec.email>" in SKILL.md.
  • Risk: Unsanitized input for the URL or user credentials could allow for command injection if the input contains shell metacharacters such as backticks or subshells.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes a URL from external agent context without sanitization, providing an attack surface for malicious instructions or characters.
  • Ingestion points: The <url> argument in SKILL.md provided by the calling skill.
  • Boundary markers: Absent. The input is used directly in a shell command and passed to a script.
  • Capability inventory: Shell execution (bash) and network requests (curl).
  • Sanitization: Absent. The skill does not verify the domain or sanitize the string before execution.
  • [DATA_EXFILTRATION]: The skill sends user-identifiable information (name and email) to any URL provided to the fetch command.
  • Evidence: curl -sf -H "User-Agent: ${NAME} ${EMAIL}" "$URL" in scripts/sec-fetch.sh.
  • Risk: Although the skill is intended for SEC domains, it does not restrict the target domain. A malicious agent could provide a URL to an attacker-controlled server, causing the skill to exfiltrate the user's name and email address via the User-Agent header.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Sep 17, 2026, 07:27 AM
Security Audit — agent-trust-hub — sec-fetch