The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes local shell commands to perform file management tasks within the user's environment.
Evidence: Uses ls "$VAULT/Inbox/" --ignore="Attachments" to retrieve a list of files to process.
Evidence: Instructs the agent to use the mv command to relocate files according to the PARA organizational principle.
[PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection as it processes untrusted data from the local file system.
Ingestion points: File names and contents are read from the $VAULT/Inbox/ directory (SKILL.md, step 1 and 5).
Boundary markers: Absent. There are no instructions defining delimiters to separate file content from the agent's internal logic.
Capability inventory: The agent can move files via mv, modify file content (frontmatter updates), and interact with other skills like memorb-ingest.
Sanitization: Absent. The skill does not describe any validation or escaping mechanisms for the data found within the triaged files.

inbox-triage