The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local script Tools/CollectSignals.ts using bun. This script is part of the skill package and performs read-only gathering of local metadata from the user's environment (e.g., directory names in ~/.claude/LIFEOS/MEMORY/WORK).
[COMMAND_EXECUTION]: The skill uses curl to send a POST request to http://localhost:31337/notify. This is a standard pattern for local voice notifications to a developer-resident agent and does not exfiltrate data to the internet.
[DATA_EXPOSURE]: The skill reads metadata from the user's work history and frustration signals (ratings). However, it includes explicit instructions in the Scan workflow to redact secrets, client names, and personal paths from any output generated for review.
[REMOTE_CODE_EXECUTION]: No remote code execution was detected. All scripts and workflows are locally defined and restricted to gathering existing metadata.
[PROMPT_INJECTION]: No malicious prompt injection or bypass patterns were detected. The skill maintains a strict permission boundary between discovery (read-only) and creation (delegated to a separate skill CreateSkill).

SuggestSkills