The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the Bash tool to perform file system operations using /usr/bin/find and /usr/bin/grep. It also executes a local Python script located at ${CLAUDE_PLUGIN_ROOT}/scripts/ring.py for task management. These operations are restricted to standard system utilities and internal scripts provided with the skill.
[INDIRECT_PROMPT_INJECTION]: The skill creates a surface for indirect prompt injection by taking natural language input from the user and transforming it into mandates for other agent sessions.
Ingestion points: The primary user-provided <problem> argument is the main ingestion point for untrusted text.
Boundary markers: The skill uses specific markdown structures and templates defined in ring-file.md and carrier.md to wrap the delegated instructions.
Capability inventory: The skill and its child agents have access to a broad set of tools including Read, Write, Edit, Bash, and Agent, which could be misused if a malicious instruction is successfully injected.
Sanitization: While the skill performs extensive logical validations (V1-V15) on arguments and file paths, it does not explicitly describe sanitization or filtering of the natural language problem description before it is passed to sub-agents.

open