The Agent Skills Directory

[PROMPT_INJECTION]: The skill stores user-provided definitions in a glossary file that is specifically intended for injection back into the agent's prompt context later. This establishes a vulnerability surface for stored indirect prompt injection.
Ingestion points: Data enters the context via the <definition> argument in the $ARGUMENTS field within SKILL.md.
Boundary markers: There are no explicit markers or safety delimiters defined in the skill for isolating this content when it is injected into prompts.
Capability inventory: The skill uses the Bash tool to execute CLI commands and the Write tool to modify the project glossary.
Sanitization: Validation is limited to checking for kebab-case format in the term name and character length limits for the definition; no sanitization against prompt injection patterns is present.
[COMMAND_EXECUTION]: The skill executes shell commands using the maestro CLI tool. It passes user-supplied arguments directly to this tool via Bash. While the skill wraps these arguments in double quotes to prevent basic shell injection, this remains a standard execution surface that relies on the security of the underlying CLI utility.

domain-add