session-search

Pass

Audited by Gen Agent Trust Hub on Oct 15, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands using aichat search and jq to query and parse local session data.
  • [EXTERNAL_DOWNLOADS]: The documentation includes instructions for installing tools via uv tool install claude-code-tools and cargo install aichat-search to handle dependencies.
  • [REMOTE_CODE_EXECUTION]: The skill provides installation commands for common developer tools, specifically fetching the official installers for uv (https://astral.sh/uv/install.sh) and Rustup (https://sh.rustup.rs) and piping them to a shell for execution.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from previous session logs, creating a potential surface for indirect prompt injection.
  • Ingestion points: Reads from JSONL files located at ~/.claude/projects/*/[session-id].jsonl (SKILL.md).
  • Boundary markers: None present; the agent is instructed to summarize the findings directly from the file content.
  • Capability inventory: The skill uses aichat for searching and jq for parsing, with the ability to summarize content for the user.
  • Sanitization: No specific sanitization or escaping of the session content is described before the summarization step.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 15, 2026, 07:08 PM