The Agent Skills Directory

[DYNAMIC_CONTEXT_INJECTION]: The skill employs the exclamation-mark-backtick syntax in SKILL.md to automatically execute the ops-speedup binary during the skill's initialization phase. This results in a system scan running without explicit user initiation when the skill is loaded.
[COMMAND_EXECUTION]: The skill executes a local binary (ops-speedup) with high-impact system capabilities, including kernel tuning (vnodes, somaxconn), network stack modification (TCP BBR), and DNS cache flushing.
[COMMAND_EXECUTION]: The skill manages system lifecycle processes, including the ability to demote daemons and terminate processes identified as high-usage, which can impact system stability if critical services are targeted.
[COMMAND_EXECUTION]: Destructive file system operations are performed in Deep and Aggressive modes, targeting directories such as the system Trash, node_modules, and Docker volumes. While the skill includes instructions for user confirmation, these actions represent significant data removal capabilities.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted system data such as process lists and directory contents while maintaining capabilities for file deletion and process termination. Evidence: 1. Ingestion points: history.jsonl and system scans. 2. Boundary markers: Absent. 3. Capability inventory: bash execution of binary for system modification and deletion. 4. Sanitization: Absent.

ops-speedup