The Agent Skills Directory

[PROMPT_INJECTION]: Indirect Prompt Injection Surface
Ingestion points: The skill reads from context/events.jsonl and context/memory.md which contain historical interaction data.
Boundary markers: No delimiters or specific instructions are provided to the agent to ignore or isolate potentially malicious instructions embedded within these log files.
Capability inventory: The skill utilizes the Bash and Read tools, enabling it to execute commands and access file contents based on its analysis.
Sanitization: No sanitization or validation of the log content is performed before it is processed for system metrics.
[COMMAND_EXECUTION]: Local System Diagnostics
The skill executes standard shell utilities including tail, wc, and ls to analyze the local environment and its diagnostics files.
It interacts with an internal framework script python3 context/kernel.py to update the active agent state, publish system events, and manage process locks (acquire_lock, release_lock).

os-health-check