The Agent Skills Directory

[COMMAND_EXECUTION]: The skill defines shell commands for log management in Stage 5, specifically using mkdir -p and mv to rotate log.md into a dated archive folder. These commands use variables (e.g., $VAULT and date-based placeholders) that may be influenced by external input.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection attack surface as it ingests and processes untrusted user data to drive automated actions and generate file content.
Ingestion points: The skill reads from memorbs/log.md, memorbs/HQ/OrbTrack/, and Daily Notes/ to gather data for analysis (SKILL.md, Stage 2).
Boundary markers: There are no explicit boundary markers or directives defined to prevent the agent from following instructions potentially embedded within the ingested log files.
Capability inventory: The skill has the capability to write to memorbs/HQ/Core/, memorbs/HQ/Belief/, and edit 000-MOC.md files, alongside shell execution for file rotation (SKILL.md, Stage 5).
Sanitization: The instructions do not specify any sanitization or validation logic for content extracted from user logs before it is used to construct filenames (slugs) or shell command arguments.

dream-studio