vibe-sunsang-onboard
Pass
Audited by Gen Agent Trust Hub on Mar 27, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes shell commands to prepare the user's workspace, including directory structure creation (
mkdir -p) and initializing version control (git init) in the data folder. - [COMMAND_EXECUTION]: It invokes a local Python script (
convert_sessions.py) located within the plugin's own directory to handle the transformation of session data. - [PROMPT_INJECTION]: The skill reads content from
CLAUDE.mdorREADME.mdfiles in external project directories to suggest a workspace classification (e.g., Builder, Explorer, Designer), creating a potential surface for indirect prompt injection. - Ingestion points: Reads project documentation from arbitrary paths found in the user's Claude Code metadata directory (
$HOME/.claude/projects/). - Boundary markers: Absent during the content extraction and parsing process.
- Capability inventory: The skill has the ability to write to the file system, create directories, and execute local scripts.
- Sanitization: The skill mitigates risks by presenting the inferred results to the user for explicit confirmation or correction via an interactive UI prompt before any data is saved or utilized for further logic.
Audit Metadata