The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The scripts/ensure-daily-skill-update.mjs script performs automated updates by running npx skills@latest update. This downloads and executes code from the npm registry to maintain the skill environment.
[COMMAND_EXECUTION]: The skill relies on executing external CLI tools and scripts. scripts/yunxiao_cli_runtime.py uses subprocess.run to invoke the official Alibaba Cloud aliyun CLI. Additionally, cross-platform launchers (run-skill-script.sh and run-skill-script.ps1) execute local Python scripts.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it reads and parses structured data (using HTML comments as markers) from Yunxiao work item descriptions via scripts/handoff_gate.py and scripts/yunxiao_cli_test_lifecycle.py. These descriptions could potentially be modified by any user with project access to influence the agent's logic, although the skill mitigates this through mandatory user confirmation (Plan mode) and SHA-256 integrity verification of documents.
[DATA_EXPOSURE]: The scripts/_auth.py and scripts/refresh_cookies.py scripts extract authentication cookies (including XSRF-TOKEN and AONE_SESSION) from the user's Chrome browser and store them in a temporary file (yunxiao_cookies.json) in the system's temporary directory to facilitate API access.
[DYNAMIC_EXECUTION]: The skill uses handoff_gate.py to fetch external documents via HTTP(S) for integrity verification (SHA-256). While it does not execute these documents, it demonstrates a capability to ingest remote content based on URIs found in work item metadata.

YunxiaoQA