The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands such as mktemp -d and node to manage its internal workspace and process candidate files. These operations are executed using absolute paths and are necessary for the skill's logic.
[PROMPT_INJECTION]: The skill processes user-inputted naming briefs and passes them to subagents, creating a surface for indirect prompt injection which is mitigated by its structured design.
Ingestion points: User interview data is written to a temporary generator-brief.md file in the orchestrator's scratch directory.
Boundary markers: The naming brief is placed within a clear template that identifies the content as a "naming brief" for the subagent.
Capability inventory: Subagents are tasked with writing files to disk; the orchestrator затем uses a local script to read these files and generate a final report.
Sanitization: The merge-candidates.mjs script uses a strict regular expression (CANDIDATE) to extract names and rationales from the subagent output, effectively ignoring any non-conforming content or instructions that might have been injected.

name-cracker