build-user-profile
Pass
Audited by Gen Agent Trust Hub on Apr 15, 2026
Risk Level: SAFEDATA_EXFILTRATIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The skill extracts the user's git identity (name and email) via
git log --format='%an <%ae>'. While this is intended for profile building, it constitutes exposure of personally identifiable information (PII) within the agent's persistent memory. - [COMMAND_EXECUTION]: The skill uses shell commands to inspect the workspace structure and read project configuration files, specifically using
ls -d */andcat *.yamlfor metadata harvesting. - [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted data. 1. Ingestion points: git commit history, README files, and project configuration files (e.g., package.json, requirements.txt). 2. Boundary markers: No delimiters or ignore instructions are used when reading external files. 3. Capability inventory: The skill has access to
Write,Edit, andBashtools. 4. Sanitization: No sanitization or validation of the ingested content is performed before it is synthesized into the profile.
Audit Metadata