The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests external data through the Research tool in the CONSUME and STEAL phases. While this represents a potential attack surface, the skill utilizes structured JSON data and atomic idea extraction which provides boundary separation and reduces the risk of the agent obeying instructions embedded in the ingested content.
[COMMAND_EXECUTION]: The workflows utilize a local curl command to send notifications to http://localhost:31337. These calls use hardcoded JSON messages and do not incorporate untrusted user input, effectively preventing command injection.
[DATA_EXPOSURE]: The skill reads and writes state information to local directories within ~/.claude/LIFEOS/. This behavior is consistent with the skill's intended operation for memory persistence and does not target sensitive system files or user credentials.

Ideate