Security audit
sqlserver-tidb-replay
Security checks across malware telemetry and agentic risk
Overview
This appears to be a real SQL replay tool, but it can run raw data-changing SQL against TiDB using high-privilege database credentials, so it should be reviewed carefully before use.
Install only if you intend to replay SQL against a controlled TiDB test environment. Do not use production credentials or the root account; review and filter the replay JSON first, back up the target database, and protect generated reports because they may contain sensitive SQL.
VirusTotal
67/67 vendors flagged this skill as clean.
Risk analysis
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
A selected log file could change, delete, or corrupt data in the target TiDB database during replay.
The replay script executes SQL strings from the replay input directly on the target TiDB connection. Because replayed logs can include INSERT, UPDATE, DELETE, DDL, or other state-changing statements, this can mutate or destroy target database data if run against the wrong environment.
with conn.cursor() as cursor:
cursor.execute(sql)Run only against an isolated test or disposable TiDB instance, inspect the generated JSON before replay, filter out DDL/destructive statements, and add backups or a dry-run/approval workflow before production-like use.
If root or another highly privileged account is used, replayed SQL can affect more data and schema than intended.
The script accepts database credentials and defaults the TiDB user to root. This is purpose-aligned for connecting to TiDB, but root is an overpowered default for a tool that replays arbitrary SQL.
parser.add_argument("--user", default=os.getenv("TIDB_USER", "root"), help="TiDB 用户")
parser.add_argument("--password", default=os.getenv("TIDB_PASSWORD", ""), help="TiDB 密码")Create a dedicated least-privilege TiDB account for the specific test database, avoid root, and ensure the registry/installation documentation clearly declares the credential requirement.
Users may need to find or create the missing collection script, which could introduce unreviewed code into the workflow.
The documentation references a PowerShell collection helper, but the provided file manifest does not include collect_xe.ps1. This is a provenance gap if users obtain or run that helper from elsewhere.
.\collect_xe.ps1 -SessionName "slow_query_capture" -ThresholdMs 1000 -OutputPath "C:\slow_logs\slow_20260412.csv"
Only run a collector script from a trusted, reviewed source, and include the referenced helper in the package or clearly document that users must supply their own.
Generated CSV/JSON/HTML reports may contain sensitive SQL text and should not be shared casually.
The conversion output stores raw SQL fragments for comparison. This is expected for reporting, but slow-query logs may contain sensitive table names, business data literals, or identifiers.
'original_raw': raw_sql[:500], # 保存原始片段用于对比
Store outputs securely, redact sensitive literals before sharing reports, and delete intermediate files when the migration test is complete.