init-mode
Pass
Audited by Gen Agent Trust Hub on Sep 6, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- Local Configuration Modification: The skill executes commands to read the user's name from
git configand modifies the repository's.gitattributesfile to set merge strategies. While functional for team synchronization, these operations interact with the local development environment and repository configuration. - Indirect Prompt Injection Surface: The skill features a data ingestion path that warrants review. It collects project descriptions and tech stacks from the user and interpolates them into agent
history.mdfiles. This creates a surface where external input flows into the agent's long-term context. - Ingestion points: Project description and tech stack inputs (SKILL.md).
- Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present when writing user input to agent context.
- Capability inventory: The skill can write to the filesystem, create directories, and update git attributes across the
.squad/directory. - Sanitization: No explicit filtering or sanitization of the user-provided project description is performed before it is added to agent memories.
- PII Protection Measures: The skill instructions include a specific safeguard to prevent the collection or storage of email addresses from git configuration. This minimizes the exposure of personally identifiable information during the team initialization process.
Audit Metadata