result-bitable-reporter

Warn

Audited by Gen Agent Trust Hub on Mar 1, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONDATA_EXFILTRATIONCREDENTIALS_UNSAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes multiple system-level commands to manage data and processes.
  • It utilizes sqlite3 to query and update the local database located at ~/.eval/records.sqlite.
  • It uses the ps utility to list processes and process.kill to terminate existing collection tasks based on their process IDs.
  • It spawns an external background process named evalpkgs for real-time data collection. The source and integrity of this binary are not provided within the skill package.
  • [DATA_EXFILTRATION]: The core functionality involves reading records from a local SQLite database and transmitting them to the Feishu Bitable API (open.feishu.cn). Although Feishu is a well-known service, the automated exfiltration of local event data to an external platform is a sensitive operation.
  • [CREDENTIALS_UNSAFE]: The skill mandates the use of sensitive environment variables, FEISHU_APP_ID and FEISHU_APP_SECRET, for API authentication. If these credentials are leaked or improperly stored in the agent's environment, it could lead to unauthorized access to the linked Feishu application.
  • [COMMAND_EXECUTION]: The skill exposes a --where flag that allows the inclusion of raw SQL fragments in database queries. While there is a rudimentary sanitization mechanism for arguments (sqlLiteral), the structure of the SQL predicate itself is user-controlled (or agent-controlled), which represents a vulnerability if untrusted data is interpolated into the command flags.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 1, 2026, 11:02 PM