The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from user responses via the AskUserQuestion tool and the $ARGUMENTS parameter. This input influences the creation of directory names, metadata schemas, and note templates. The instructions lack explicit sanitization or boundary markers to prevent potentially malicious strings in user descriptions from affecting derived configurations. \n- [COMMAND_EXECUTION]: The skill utilizes the Bash tool to perform workspace operations, including directory creation (mkdir), executing a local validation script (validate-kernel.sh), and updating semantic search indexes via the qmd CLI tool. These commands are integral to the skill's infrastructure management. \n- [DYNAMIC_EXECUTION]: The skill generates new instructions for the agent by creating a 'domain guide' (SKILL.md) in the agent's configuration directory. It also creates new Markdown templates with embedded metadata schemas. These artifacts are dynamically assembled from information gathered during the conversation phase.

add-domain