The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill implements a plugin-based architecture that uses importlib.import_module in scripts/address_book/__init__.py and scripts/acl_policy/__init__.py to dynamically discover and load address book and ACL policy handlers from the local file system. While this is a standard extensibility pattern, loading code from computed paths is flagged as a potential risk.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external Alibaba Cloud APIs and local Excel backup files to perform analysis and restore operations. Maliciously crafted data within these sources could potentially influence the agent's behavior during the reporting phase.
Ingestion points: External data enters the system through call_api in scripts/core.py and is read from Excel workbooks in scripts/address_book/base.py and scripts/diff_restore.py.
Boundary markers: The skill lacks explicit delimiters or instructions to ignore embedded commands within the processed data during analysis tasks like hit, dup, or audit.
Capability inventory: The skill possesses capabilities to perform network requests via requests, write to the local file system using openpyxl, and modify cloud configurations through Alibaba Cloud APIs.
Sanitization: While the skill performs data normalization for comparison in scripts/diff_restore.py, it does not implement specific sanitization to prevent data from being interpreted as instructions by the LLM.
[SAFE]: Credential management follows security best practices by utilizing the official Alibaba Cloud Credentials SDK provider chain, avoiding hardcoded secrets, and explicitly instructing the agent never to dump credential files or environment variables. Network operations are restricted to official vendor-owned domains (aliyuncs.com).

alibabacloud-cloud-firewall-acl-manager