iros-registry-automation
Pass
Audited by Gen Agent Trust Hub on Oct 1, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill fetches tools and scripts from public repositories, specifically cloning the
challengekim/iros-registry-automationrepository from GitHub and utilizing the@nomadamas/k-skillpackage via the npm registry. These dependencies are standard for technical automations and the GitHub reference is pinned to a specific commit hash for integrity. - [COMMAND_EXECUTION]: The workflow relies on shell commands to set up the environment, including package installation via pip, Playwright browser setup, and executing the automation scripts via npx and Python.
- [INDIRECT_PROMPT_INJECTION]: The skill processes legal documents (PDFs) from IROS, which is an untrusted data ingestion surface. This is mitigated through local processing and a dedicated script (
scripts/iros_pdf_summary.py) that implements Resident Registration Number (RRN) masking via regex and limits output to objective facts. Ingestion point: local PDF files downloaded from IROS. Boundary markers: instructions use dedicated temporary directories with restricted permissions (chmod 700). Capability inventory: file system access and subprocess calls. Sanitization: automated PII masking logic. - [DYNAMIC_EXECUTION]: The skill includes Python code to dynamically generate configuration files (
config.json) and process document text at runtime. These operations are conducted locally within a temporary workspace to ensure sensitive information does not persist in the repository environment.
Audit Metadata