read-logs
Pass
Audited by Gen Agent Trust Hub on Jun 19, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTIONREMOTE_CODE_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill exposes an indirect prompt injection surface by ingesting untrusted log data from the application run.
- Ingestion points: Data is read from
/tmp/flutter-*.logwithin the Step 2 bash block inSKILL.md. - Boundary markers: Absent; log content is interpolated directly into the model context without delimiters to distinguish it from instructions.
- Capability inventory: The skill has access to shell execution (
git,grep,sed,find) and Python execution (python3 -c). - Sanitization: Performs ANSI escape code removal and JSON extraction, but does not sanitize the text content of the logs for malicious instructions.
- [DATA_EXFILTRATION]: The skill accesses sensitive application logs that may contain credentials (tokens) and PII. While it warns the user of this behavior, it facilitates the exposure of this sensitive data to the model.
- [REMOTE_CODE_EXECUTION]: Employs
python3 -cto execute an inline Python script for extracting app output from JSON-framed Debug Adapter Protocol (DAP) logs. - [COMMAND_EXECUTION]: Uses multiple shell scripts to resolve repository paths via Git, check file freshness using
stat, and detect or modify editor configuration files (.vscode/settings.jsonand.zed/tasks.json).
Audit Metadata