The Agent Skills Directory

[SAFE]: The skill implements a coordinated batch-rename pattern using parallel sub-agents. It demonstrates security awareness through several design choices: 1. Boundary Constraints: The prompt template explicitly requires the user to define what should not be renamed, reducing the risk of unintended modifications to sensitive fields such as API keys or external imports. 2. Verification Wave: It includes a multi-step process where core changes are verified via Grep and CI tools (devrun) before completion. 3. Isolated Execution: By using one agent per file, it limits the blast radius of any individual failed or misinterpreted rename operation. 4. Surface Analysis: While the skill processes untrusted file data (Ingestion points: source/test files; Capabilities: Grep, Task, Edit, devrun), the risk of indirect prompt injection is mitigated by the narrow mechanical nature of the task, the use of explicit boundary markers in the agent prompts, and post-execution verification steps.

rename-swarm