parallel-web-extract
Pass
Audited by Gen Agent Trust Hub on Nov 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to fetch and process content from arbitrary external URLs, creating an attack surface for indirect prompt injection attacks.
- Ingestion points: Untrusted data enters the agent context through web content retrieved via the
parallel-cli extractcommand as defined inSKILL.md. - Boundary markers: The instructions do not specify any delimiters or boundary markers to separate the retrieved web content from the agent's internal instructions.
- Capability inventory: The agent has access to the
Bashtool (specificallyparallel-cli) and file system access to save output in the/tmp/directory. - Sanitization: No sanitization, validation, or filtering of the extracted content is mentioned before it is processed or returned to the user.
- [COMMAND_EXECUTION]: The skill dynamically constructs shell commands using input provided through
$ARGUMENTSand agent-generated filenames. - Evidence:
SKILL.mdinstructs the agent to "Construct arguments directly from the requested URLs" and to substitute placeholders "inline" into the command. - Mitigation: The skill includes explicit security instructions to use quoted positional arguments and avoid the use of
evalto minimize the risk of shell command injection.
Audit Metadata