hyper-extract
Pass
Audited by Gen Agent Trust Hub on Oct 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external documents to extract structured knowledge, which introduces an attack surface for indirect prompt injection. Malicious instructions embedded in processed documents could attempt to influence the agent's extraction logic or subsequent actions.
- Ingestion points: Input files (PDF, Markdown, etc.) parsed by the
he parsecommand as described inSKILL.mdandreferences/upstream.md. - Boundary markers: The skill utilizes templates to define extraction structure but does not implement explicit delimiters or 'ignore instructions' markers to isolate the raw document content within the agent's context.
- Capability inventory: The skill possesses the capability to execute shell commands via the
Bashtool, write output files, and interact with remote LLM providers (SKILL.md,references/upstream.md). - Sanitization: The skill's provided scripts (e.g.,
preflight.py) do not perform content sanitization or instruction filtering on the documents before they are processed by the CLI. - [COMMAND_EXECUTION]: The skill's core functionality relies on executing the
heCLI through theBashtool. While this is the intended purpose, it represents a high-privilege capability. The skill mitigates risks by providing detailed instructions on safe command shapes and advising against exposing API keys in shell commands.
Audit Metadata