feishu-evolver-wrapper
Warn
Audited by Gen Agent Trust Hub on Feb 19, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- COMMAND_EXECUTION (MEDIUM): Several scripts utilize child_process.exec and execSync to manage process lifecycles and run maintenance tasks. exec_cache.js provides a generic wrapper for executing shell commands with a 1-minute cache. issue_tracker.js executes local scripts with arguments derived from environment variables like OPENCLAW_MASTER_ID. self-repair.js executes git commands such as rebase, merge, and fetch.
- EXTERNAL_DOWNLOADS (MEDIUM): The skills_monitor.js script automatically executes 'npm install' for other skills discovered in the peer directory. This involves downloading and potentially executing code from the NPM registry. self-repair.js also performs 'git fetch' from origin.
- DATA_EXFILTRATION (LOW): The skill is designed to send internal logs and evolution events to Feishu. It reads sensitive files like feishu_token.json and environment variables. Mitigation is present via feishu-helper.js, which uses a scanForSecrets function with regex patterns to block known credential formats for Anthropic, GitHub, Slack, and generic Private Keys.
- PROMPT_INJECTION (LOW): visualize_dashboard.js and export_history.js process log files (events.jsonl, mad_dog_evolution.log) that may contain untrusted data from agent-driven activities. This data is parsed and formatted into Markdown for reporting without extensive sanitization against indirect injection patterns.
Audit Metadata