The Agent Skills Directory

[SAFE]: The skill is purely informational, detailing the use of state triggers and function registration primitives. No security threats or malicious patterns were identified in the content.
[COMMAND_EXECUTION]: No shell commands or script execution mechanisms are present in the documentation.
[EXTERNAL_DOWNLOADS]: The skill references local files for implementation details and does not contain commands to download or execute code from remote servers.
[PROMPT_INJECTION]: (Indirect) The skill describes a pattern for observing state changes (new_value, old_value). While this defines a data ingestion surface where external state data could theoretically contain instructions, the documentation itself does not include any injection techniques.
Ingestion points: state::set, state::update, and state::delete events described in SKILL.md.
Boundary markers: Not specified in the architectural overview.
Capability inventory: Primitive registration (registerFunction, registerTrigger); no dangerous OS or network capabilities are invoked.
Sanitization: Not explicitly addressed in the pattern guidance.

state-reactions