wb-xhs-topic-bank
Fail
Audited by Gen Agent Trust Hub on Aug 2, 2026
Risk Level: HIGHREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The 'Runtime bootstrap and update gate' section in SKILL.md instructs the agent to 'automatically install or update a missing or outdated dependency to the latest stable supported version' before using external programs or scripts. This involves fetching and integrating external code at runtime that is not statically defined or verified within the skill itself.
- [COMMAND_EXECUTION]: The skill requires the agent to 'run its diagnostic' on newly installed or updated dependencies. This execution of unverified code at runtime is a dangerous pattern that can lead to arbitrary command execution if a dependency is compromised.
- [PROMPT_INJECTION]: The skill identifies a vulnerability surface for indirect prompt injection. It processes user-supplied positioning, keywords, and persona details to generate topic banks without using boundary markers or sanitization steps. 1. Ingestion points: User-provided positioning and niche data processed in the 'E — 可执行步骤' section. 2. Boundary markers: Absent; the skill does not use delimiters or instructions to ignore embedded commands within user inputs. 3. Capability inventory: The agent is instructed to perform package installation and diagnostic execution. 4. Sanitization: Absent; user inputs are interpolated directly into content generation prompts without filtering.
Recommendations
- AI detected serious security threats
Audit Metadata