clawhub-moderation
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses the
bun run admincommand to perform administrative operations on the ClawHub platform. These operations include high-privilege actions like hard-deleting skills (skills hard-delete), banning users (users ban), and sending emails via an internal server.\n- [INDIRECT_PROMPT_INJECTION]: The skill accepts arbitrary user input for command arguments such as reasons, notes, and identifiers, which are then passed to the local shell tool.\n - Ingestion points: User-provided strings for placeholders like
<slug>,<reason>,<handleOrId>,<note>, and<csv>inSKILL.md.\n - Boundary markers: The skill mandates confirmation by the user before any write operation and suggests the use of dry-run tokens for destructive actions.\n
- Capability inventory: The skill has the capability to execute administrative shell commands that interact with the ClawHub API and production noreply email provider.\n
- Sanitization: The instructions recommend wrapping arguments in quotes (e.g.,
--reason "<reason>") but do not provide specific guidance for escaping shell metacharacters or validating the integrity of user-supplied paths like<csv>.
Audit Metadata