The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection vulnerabilities.
Ingestion points: It retrieves untrusted data from external sources via issue show, issue list, and issue comment list subcommands defined in SKILL.md.
Boundary markers: There are no specific delimiters or instructions used to prevent the agent from mistakenly following commands embedded in the work item titles, descriptions, or comments.
Capability inventory: The agent can execute a wide range of codearts-cli commands (create, update, delete) and write image files to the local filesystem using the --download-images flag.
Sanitization: While the codearts-cli performs magic bytes validation on image downloads, no sanitization or escaping is mentioned for the textual content of retrieved issues or comments.
[COMMAND_EXECUTION]: All management functionality is implemented through the execution of the codearts-cli binary.
The skill facilitates project management tasks including creating, updating, and querying resources on the Huawei Cloud CodeArts platform.
It uses standard piping to jq to parse local configuration data, such as userId, from the CLI's own output.

codearts-issue