The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The helper script scripts/skill-agent-table.mjs executes external code downloaded from the NPM registry at runtime using npx -y skills@1.5.3. This pattern executes code from an external source that is not part of the skill's own distribution or a widely recognized trusted vendor.
[COMMAND_EXECUTION]: The skill is designed to have the agent execute administrative shell commands using the skills CLI, including operations like install, update, remove, and repair, which can modify the system's software environment.
[DYNAMIC_EXECUTION]: The script scripts/skill-agent-table.mjs uses node:child_process's spawnSync to execute the npx command and the skills CLI, representing a dynamic execution pattern where the agent invokes shell processes via a helper script.
[INDIRECT_PROMPT_INJECTION]: The skill processes JSON data retrieved from the skills CLI and renders it into a table format for the agent's consumption. This creates a vulnerability surface for indirect prompt injection.
Ingestion points: Data enters the agent's context through scripts/skill-agent-table.mjs which reads the output of npx skills list --json.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to ignore potentially malicious instructions embedded in the skill metadata (e.g., skill names or paths) processed by the script.
Capability inventory: The skill has the capability to write to the file system and execute shell commands via spawnSync in scripts/skill-agent-table.mjs and instructions in SKILL.md to use the CLI for installation and removal.
Sanitization: The script performs no escaping or sanitization of the skill.name or skill.path fields before interpolating them into the output table.

vp-skills