marketing-refresh
Warn
Audited by Gen Agent Trust Hub on Mar 23, 2026
Risk Level: MEDIUMDATA_EXFILTRATIONPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [DATA_EXFILTRATION]: Scans sensitive locations like
$HOME/personal/*and$HOME/work/*during project discovery, exposing private files to the agent context. - [PROMPT_INJECTION]: Vulnerable to indirect prompt injection. Ingestion points: local project files (READMEs, manifests) and web search results (competitor pages, forums, and reviews). Boundary markers: none present in instructions. Capability inventory: file system access, network access (web search), and agent tool invocation for parallel tasks. Sanitization: no evidence of content escaping or validation before processing.
- [COMMAND_EXECUTION]: The
$ARGUMENTSvariable is used for project targeting, which can lead to command injection if processed in a shell context without proper sanitization. - [DATA_EXFILTRATION]: Automatically records metadata to
~/.claude/projects/skill-telemetry.md, which is a form of persistent local data collection.
Audit Metadata