session-scout
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses PowerShell and Bash scripts to execute system commands including
wsl.exe,find, andgrepfor session artifact discovery and parsing.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted local session logs which serve as an ingestion point for external data into the agent's context during the auditing process.\n - Ingestion points: The scripts in the
scripts/directory read.jsonland.logfiles from~/.claude/projects/and~/.local/share/opencode/log/.\n - Boundary markers: None identified; extracted data is displayed as a table or written to a local CSV file.\n
- Capability inventory: The skill possesses shell execution capabilities for file system discovery and local file-write capabilities for generating audit reports.\n
- Sanitization: The scripts use targeted regular expressions to extract only specific metadata fields such as
cwdandSessionId, which mitigates the risk of importing malicious instructions.
Audit Metadata