The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the curl tool to send JSON payloads to http://localhost:31337/notify. This is used to trigger local voice notifications for the user. Since the destination is restricted to localhost and serves a UI notification purpose, it is considered safe functional behavior.
[COMMAND_EXECUTION]: The skill implements execution logging by appending JSON records to ~/.claude/LIFEOS/MEMORY/SKILLS/execution.jsonl using the echo command. This practice is part of the skill's management of its own history within the user's home directory.
[COMMAND_EXECUTION]: The workflow documentation in DesignExperiment.md instructs the agent to run local TypeScript scripts using bun run. The paths provided (~/.claude/skills/Evals/Tools/EvalRunner.ts) point to tools within the same controlled local framework.
[PROMPT_INJECTION]: The skill identifies an ingestion point for local user customizations that can override default behavior.
Ingestion points: The directory ~/.claude/LIFEOS/USER/CUSTOMIZATIONS/SKILLS/Science/ is checked for PREFERENCES.md and other configuration files.
Boundary markers: The instructions do not specify the use of delimiters or 'ignore' instructions when processing the content of these local configuration files.
Capability inventory: The skill environment allows for shell command execution (curl, echo, bun) and file system access within the ~/.claude hierarchy.
Sanitization: There is no explicit requirement for the agent to sanitize or validate the content of the customization files before applying them.

Science