The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides templates for and executes shell commands (Bash and PowerShell) to launch external CLI tools such as claude and opencode for model orchestration.
[EXTERNAL_DOWNLOADS]: Fetches technical reports and token analysis documentation from the official GitHub repository of the skill author (ruan-cat/11comm). These are vendor-provided resources and are handled as trusted documentation.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it processes logs, results, and files generated by delegated sub-agents. The design mitigates this risk by explicitly instructing the primary agent to perform mandatory independent verification of all changes. 1. Ingestion points: Agent reads result.json, result.jsonl, and execution-log.md. 2. Boundary markers: 'Task Packet' (context-packet.md) defines constraints. 3. Capability inventory: Delegated tools have filesystem and command capabilities. 4. Sanitization: Instructions require manual re-verification of all outputs.
[DYNAMIC_EXECUTION]: The skill involves the generation and execution of temporary shell scripts (PowerShell and Bash) to facilitate the model delegation process. These scripts are created from internal templates using agent-managed variables and do not incorporate unsanitized remote input into their logic.

use-other-model