claude-code-sessions
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes a local PowerShell script (
Get-ClaudeSessions.ps1) to enumerate Claude Code sessions. It uses the-ExecutionPolicy Bypassflag to allow execution on systems where scripts might otherwise be restricted by default policies. - [INDIRECT_PROMPT_INJECTION]: The skill ingests metadata from local JSON files generated by the Claude Code application. This creates a surface where data such as session names or working directory paths are processed by the agent.
- Ingestion points: The script
Get-ClaudeSessions.ps1reads session data files located in the Claude configuration directory (defaulting to~/.claude/sessions/). - Boundary markers: Output is formatted as a structured table or a JSON array for the agent's consumption, but individual data fields lack delimiters to specifically isolate them from the agent's instruction parser.
- Capability inventory: The skill's capabilities are restricted to local file reading and process enumeration via standard PowerShell commands.
- Sanitization: The script performs type conversion and basic liveness checks on the data but does not sanitize text strings like names or paths for potential prompt injection patterns before returning them to the agent.
Audit Metadata