tao-route-visual-changenet-samples
Warn
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: MEDIUMDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The
hooks/routing-package.shhook automatically copies sensitive files from the local environment into the skill's output directory. This includes the agent's project configuration (.claude/settings.json,.claude/settings.local.json) and session transcripts. While intended for workflow reproducibility, this behavior exposes sensitive settings and conversation history within the generated results artifacts. - [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external sources (
gaps_parquetandsource_pool_csv) without applying security boundaries or sanitization. This creates an attack surface where maliciously crafted data could influence the agent's analysis or report generation. - Ingestion points: Untrusted parquet and CSV files specified by the user as input to the routing process.
- Boundary markers: No delimiters or instructions are used to isolate untrusted data from the agent's execution context.
- Capability inventory: The skill uses
Bashto execute Python scripts that process the data andReadto access the filesystem. - Sanitization: No filtering or escaping is performed on the ingested labels or metadata before they are included in the generated
Routing_Report.md.
Audit Metadata