The Agent Skills Directory

[COMMAND_EXECUTION]: The Python script scripts/openharmony_ci.py executes the oh-gc CLI tool via subprocess.run to retrieve PR comments. The command is invoked using a structured list of arguments and a controlled environment (XDG_CACHE_HOME), which minimizes command injection risks.
[EXTERNAL_DOWNLOADS]: The skill performs HTTP GET requests to dcp.openharmony.cn and cidownload.openharmony.cn to fetch CI metadata and build logs. These are project-specific domains for OpenHarmony and align with the skill's stated purpose.
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: Data is ingested from external GitCode PR comments and CI logs.
Boundary markers: No specific delimiters or safety instructions are identified that isolate external data from the agent's prompt context.
Capability inventory: The skill possesses capabilities for command execution, network requests, and local file writing.
Sanitization: Event IDs are validated using a hexadecimal regex, but the content of retrieved logs is not sanitized for potential embedded prompt instructions.

openharmony-ci