secondbrain-search-init
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFEPERSISTENCEDYNAMIC_EXECUTIONCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTIONNO_CODE
Full Analysis
- [PERSISTENCE]: The skill configures a 'PostToolUse' hook within '.claude/settings.local.json'. This mechanism ensures that a Python script ('search-index-update.py') is automatically executed whenever the agent performs 'Write' or 'Edit' operations, allowing the search index to persist in a synchronized state with the file system.
- [DYNAMIC_EXECUTION]: Implements a runtime execution pattern where shell commands and Python scripts are invoked automatically based on agent tool-usage events.
- [COMMAND_EXECUTION]: Utilizes system commands such as 'qmd index', 'which qmd', and directory management tools to initialize and maintain the search infrastructure.
- [REMOTE_CODE_EXECUTION]: The indexing process involves downloading large machine learning models (specifically 'embedding-gemma-300M' and 'qwen3-reranker') to the local machine (~1.5GB) for generating vector embeddings.
- [INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest and index local documents from the 'docs/' directory.
- Ingestion points: Reads all files within the 'docs/' path to build search chunks.
- Boundary markers: The configuration does not specify explicit delimiters or 'ignore' instructions for the indexer when processing these files.
- Capability inventory: The skill has the ability to execute shell commands ('qmd') and Python scripts via hooks.
- Sanitization: There are no instructions for sanitizing or escaping the content of the indexed documents before they are processed by the semantic search tool.
Audit Metadata