cold-start-interview
Pass
Audited by Gen Agent Trust Hub on Jul 3, 2026
Risk Level: SAFEDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The skill is designed to ingest and analyze highly sensitive legal documents, including signed commercial agreements, escalation matrices, and negotiation playbooks. While this is necessary for the skill's stated purpose of building a personalized practice profile, it creates a significant sensitive data exposure surface within the agent's processing context.
- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection during the 'seed document' analysis phase. Maliciously crafted agreements could theoretically contain instructions designed to bias or corrupt the extracted 'playbook positions' which are then written to a persistent configuration file (
CLAUDE.md) that influences all other plugin skills. - Ingestion points: Specifically in Part 2 (The playbook) and Part 4 (Seed documents), where the agent reads provided file contents, Google Drive links, or CLM record IDs.
- Boundary markers: The instructions do not define specific delimiters or instructions to the agent to ignore hidden commands within the ingested legal documents.
- Capability inventory: The skill has permissions to write to the user's home directory (
~/.claude/plugins/config/...) and can interact with integrated tools such as Slack, e-signature platforms, and document storage via MCP tool calls. - Sanitization: No sanitization logic or content validation is specified for the data extracted from seed documents before it is saved into the practice profile.
Audit Metadata