taxonomy-builder
Fail
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: HIGHPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- Indirect Prompt Injection (HIGH): The skill is vulnerable to malicious instructions embedded within the untrusted paper data it processes. \n
- Ingestion points: The skill ingests data from
papers/core_set.csvandpapers/papers_dedup.jsonl(Workflow step 1). \n - Boundary markers: There are no delimiters or instructions to ignore embedded commands when the agent 'skims' or 'clusters' the data. \n
- Capability inventory: The skill allows the agent to execute a local Python script (
scripts/run.py) and write to the file system (outline/taxonomy.yml). \n - Sanitization: No sanitization or validation of the input data is performed before it is processed by the agent. \n- Command Execution (LOW): The skill documentation includes commands to run a local helper script (
scripts/run.py). While the script is internal to the skill, this execution capability increases the potential impact of an injection attack.
Recommendations
- AI detected serious security threats
Audit Metadata