The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute a variety of shell utilities to query and display email data.
Evidence: Uses mu find, mu view, mu info, neomutt, nytid, sed, sort, uniq, wc, and column to process and display information.
[DATA_EXFILTRATION]: The skill accesses sensitive local email storage and private configuration files.
Evidence: Reads email data from /home/dbosk/mail/kth and configuration from /home/dbosk/.muttrc.scores.
[PROMPT_INJECTION]: The skill processes untrusted email subjects and bodies, which is a common vector for indirect prompt injection attacks.
Ingestion points: mu find and mu view (SKILL.md) ingest external email content into the agent's context.
Boundary markers: There are no explicit delimiters or instructions to ignore embedded commands within the processed email content.
Capability inventory: The skill has access to shell utilities (mu, neomutt, nytid) which could be misused if malicious instructions are followed.
Sanitization: The instructions include basic shell escaping for command construction but lack sanitization or validation for the ingested email content itself.

inbox-info