ai-writing-detector
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-provided text through the
scripts/detect.jsutility, which creates a surface for indirect prompt injection where malicious instructions could be embedded in the text being scanned. - Ingestion points: The utility reads input from standard input or local files via the
--fileargument as described inSKILL.md. - Boundary markers: There are no explicit delimiters or markers to separate the untrusted input from the agent's instructions within the script execution flow.
- Capability inventory: The skill leverages the agent's capability to read files (
fs.readFileSync) and display output to the user. - Sanitization: The
SKILL.mdincludes a robust "Anti-Patterns" section that explicitly warns the agent against treating external content, logs, or tool responses as trusted instructions, mitigating the risk of the agent obeying embedded commands. - [COMMAND_EXECUTION]: The skill's instructions in
SKILL.mddirect the agent to execute shell commands (e.g.,node scripts/detect.js) with user-provided text. This relies on the execution platform's ability to safely handle shell interpolation and prevents arbitrary command injection.
Audit Metadata