The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill instructs the agent to dynamically locate a maintenance script (reconcile.mjs) and execute it using the Node.js runtime. The execution path is determined at runtime by searching through several potential installation directories, which constitutes dynamic execution of local scripts.\n- [COMMAND_EXECUTION]: The agent is directed to execute shell commands, specifically the Node.js runtime for script execution and the npx skills command for package management and updates.\n- [REMOTE_CODE_EXECUTION]: The skill utilizes the npx skills add command to download and install packages from remote GitHub and Gitee repositories. This process involves fetching and potentially executing external code from the author's repositories.\n- [EXTERNAL_DOWNLOADS]: The reconciliation process involves fetching configuration files and package content from remote hosting services including GitHub and Gitee. These operations target repositories associated with the skill's author.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from a remote index.json file, including changelogs and notes, which are then relayed to the user, creating a surface for instruction injection if the remote source is compromised.\n
Ingestion points: Remote index.json and upstream SKILL.md files downloaded from the repository.\n
Boundary markers: There are no explicit delimiters or safety instructions provided to the agent to isolate the content of fetched files from the core instruction set.\n
Capability inventory: The skill has the capability to modify the local file system (archiving and moving directories) and execute shell commands via Node.js.\n
Sanitization: The documentation does not specify any sanitization or validation logic for the remote metadata before it is processed by the agent.

dby-update