j5ik2o-takt-workflow-builder
Pass
Audited by Gen Agent Trust Hub on Jul 15, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [DATA_EXFILTRATION]: The audit-security workflow defined in audit-security.yaml is designed to read all project files to identify vulnerabilities. This broad filesystem access is documented as a core function for auditing purposes and is conducted using local read tools.
- [EXTERNAL_DOWNLOADS]: The research-dig.md instruction allows the agent to download non-executable data files (CSV, JSON, Excel) from trusted sources. Security guidelines within the facet require the agent to verify source reliability and strictly forbid downloading executable content.
- [COMMAND_EXECUTION]: The skill uses git and gh command-line tools to manage development workflows. Facets such as gather-review.md incorporate defensive instructions to prevent shell command injection by forbidding the direct interpolation of raw user input into command strings.
- [PROMPT_INJECTION]: Untrusted data enters the agent context through pull request descriptions (gather-review.md) and external research data (research-dig.md). The 'Faceted Prompting' architecture acts as a boundary marker mechanism to separate instructions from data. Capabilities include shell execution and file system access. Sanitization is present via explicit instructions to avoid shell interpolation and distinguish speculation from fact.
- [SAFE]: The skill includes high-quality security knowledge facets (knowledge/security.md) that provide standard criteria for detecting injection risks, authentication failures, and data protection issues in the code it audits.
Audit Metadata