AI Agent Hub
Back to plugins
🧰

dsh-tool-sqlite

Web Tools Updated 2026.08.26

Run the following command in DeepSeek Harness:

dsh plugin install WODE25500/dsh-tool-sqlite

Paste the following prompt into your AI chat to install this plugin:

Run dsh plugin install WODE25500/dsh-tool-sqlite to install this plugin in DeepSeek Harness. Source: https://github.com/WODE25500/dsh-tool-sqlite

About this plugin

When an Agent session encounters .db or .sqlite files, the usual workaround is switching to bash, running sqlite3, and pasting raw text back into context. dsh-tool-sqlite turns that into a single function call: the Agent lists database files, inspects table schemas, and runs read-only queries, receiving structured columns and rows directly.

Built on Node 22+ built-in node:sqlite with zero third-party dependencies, the plugin registers five tools: sqlite_list scans the workspace (2 levels deep) for all database files; sqlite_tables lists every table and view; sqlite_schema reveals column types, primary keys, and defaults; sqlite_summary produces per-column row counts, distinct values, min/max/avg stats so you can understand data distribution before a full SELECT *; sqlite_query executes a read-only SQL statement and returns JSON. The recommended three-step workflow—summary, schema, then targeted query—dramatically reduces the tokens pulled into context.

On the safety side, every database handle is opened with readOnly: true so write statements are rejected at the engine level; a statement whitelist only allows single statements starting with SELECT, WITH, PRAGMA, EXPLAIN, or VALUES; all path parameters must resolve within the workspace or an error is raised; output is capped at 100 rows by default and 500 rows hard, with an explicit truncation note. This makes the plugin ideal for any developer or team who wants an Agent to safely read local databases—crawled snapshots, app data, export files—without risking accidental writes or context bloat.

Use Cases

  • When an Agent session hits .db or .sqlite files, list tables and run read-only queries in one call without switching to bash sqlite3
  • Use summary to check data distribution, schema to confirm column types, then a targeted query instead of SELECT * into context
  • Safely read file-based SQLite databases like crawled snapshots, export files, or local app data within the workspace

Best For

  • Developers who need Agents to safely read local SQLite databases without risk of accidental writes
  • Automation builders processing file-based databases like crawl results or export snapshots
  • Teams wanting lightweight, zero-dependency database query integration