The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input from Gitee comments, which creates a vulnerability surface for instruction injection.
Ingestion points: The skill fetches comment bodies from the Gitee platform via gitee api -p "/repos/{owner}/{repo}/comments".
Boundary markers: While it uses metadata tags like branch: and location: to structure data, it lacks explicit delimitation or instructions to ignore embedded commands within the user-provided comment text.
Capability inventory: The skill has access to the Bash tool, allowing it to execute git commands and network requests via the Gitee CLI.
Sanitization: It includes logic to handle shell quoting and multi-line inputs using temporary files to prevent command injection, but it does not sanitize the semantic content of the comments for instruction overrides.
[EXTERNAL_DOWNLOADS]: The skill suggests the installation of the official Gitee CLI package (@gitee/gitee-cli) from the npm registry for its core functionality.
[COMMAND_EXECUTION]: The skill utilizes Bash to perform repository discovery (git remote get-url), branch management (git branch --show-current), and Gitee API interactions. It includes a host check to ensure operations are restricted to gitee.com domains.

gitee-comments