The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides shell command patterns for source control operations using the svn CLI. These include instructions for writing commit messages to a temporary file (/tmp/svn_msg.txt) and using svn commit with specific encoding flags to prevent character corruption.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and act upon data from external sources such as GitHub Pull Requests and Issue descriptions. Ingestion points: The skill frontmatter and introductory text specify that the skill is used when applying changes from a GitHub PR URL or issue content provided by the user. Boundary markers: The skill mandates the use of update-begin and update-end comments to wrap all code changes, which acts as a structural boundary for the modified content. Capability inventory: The skill modifies local code files, updates project-specific log files, and executes SVN shell commands as detailed in sections I.1 through I.3. Sanitization: There are no explicit content sanitization or validation steps described for the external PR/issue data, relying instead on structural code markers.

jeecg-dev