safety-kb-query（安全知识库查询）

Safety KB Query — 安全生产法规知识库查询工具

Overview

This skill provides unified, schema-adaptive query access to the safety-review knowledge base (SQLite database). It eliminates the need to manually write connection scripts, guess field names, or handle encoding issues every time.

Database location: ~/.openclaw-autoclaw/skills/safety-review/db/knowledge.db

When to Use This Skill

• User asks about whether a standard/regulation exists in the knowledge base
• User needs to search for regulations by keyword, number, or title
• User wants to compare a list of standards against what's in the database (gap analysis)
• User needs to view clause content of a specific regulation
• User requests statistics or health checks on the knowledge base
• Any task involving the safety-review SQLite database

Trigger phrases (Chinese): 查知识库、搜法规、标准有没有、对比标准、查条款、数据质量、法规检索

How to Use

Prerequisites

1. Detect available Python command first:

   python --version

Use python or python3 based on availability.

1. The knowledge base path defaults to:

   C:\Users\13503\.openclaw-autoclaw\skills\safety-review\db\knowledge.db

Override via environment variable KB_PATH if needed.

Core Commands

All commands are executed via scripts/kb_query.py. Output is JSON for programmatic consumption.

1. Search Regulations (search)

Find regulations matching a keyword across multiple fields.

python scripts/kb_query.py search <keyword> [--mode fuzzy|exact] [--limit N]

Example scenarios:

• "看看 GB16423 有没有" → search GB16423
• "搜一下紧急避险相关的" → search 紧急避险
• "精确查找 AQ/T 2033-2023" → search "AQ/T 2033-2023" --mode exact

2. Check Standard Existence (check) ⭐ Most Useful

Given a list of standards, report which exist and which are missing — the primary gap analysis tool.

python scripts/kb_query.py check <standard1> <standard2> <standard3> ...

Output fields per result:

• found: boolean
• id: regulation ID (if found)
• document_number: stored document number
• title: regulation title (truncated)
• status: current status
• text_length: character count of full_text
• clause_count: number of associated clauses

Example:

python scripts/kb_query.py check "GB 16423" "AQ/T 2033" "AQ 2034" "国发[2010]23号"

3. Regulation Details (info)

Get comprehensive info about a specific regulation by ID.

python scripts/kb_query.py info <regulation_id>

Returns all columns from regulations table + clause count + linked std_registry entry (if any).

4. Query Clauses (clauses)

Retrieve clauses of a regulation, with optional filtering.

python scripts/kb_query.py clauses <regulation_id> [--filter <keyword>]

5. Database Statistics (stats)

Overview of knowledge base contents.

python scripts/kb_query.py stats

Returns: total regulations, total clauses, total books, status breakdown, domain distribution.

6. Schema Inspection (schema)

Auto-detect and print table structures (useful when DB schema changes).

python scripts/kb_query.py schema

7. Data Quality Check (conflicts) ⭐ Important

Detect potential data quality issues:

• Empty content: Regulations with NULL or very short full_text (<100 chars)
• No clauses: Regulations without any associated clause records
• Title/doc_no mismatch: Document number suggests a topic but title doesn't match (e.g., ID:94 containing "粮油加工" instead of mining safety content)

python scripts/kb_query.py conflicts

Workflow Integration

Typical Gap Analysis Workflow (for courseware/training materials)

When user provides a document that references standards and asks "which ones are missing":

1. Extract referenced standards from the user's document
2. Run check command with the extracted list
3. Present results as a table showing: ✅ Found / ❌ Missing / ⚠️ Data Issue
4. If data issues found, run conflicts for deeper analysis
5. Recommend next steps (import missing ones via safety-kb-import)

Typical Search Workflow

1. Run search with user's keywords
2. If results found, use info to get details on relevant entries
3. If user needs clause-level detail, use clauses
4. Present findings in readable format (table, summary)

Known Limitations

• Read-only: This skill only queries; use safety-kb-import for writing
• Encoding: Script uses UTF-8 output; ensure terminal supports it
• Fuzzy search performance: On large databases (>2000 records), LIKE queries may be slower; use --mode exact when possible
• Clause truncation: clauses command truncates content to 200 chars for readability; use info to get full text

Relationship with Other Skills

Troubleshooting

Version History

• 1.0.0 (2026-04-25): Initial release with 7 commands (search, check, info, clauses, stats, schema, conflicts)