The Agent Skills Directory

[PROMPT_INJECTION]: The skill documentation includes instructions to override safety protocols. Specifically, it mentions that the --force flag can be used to bypass safety gates during workflow execution.
Evidence: The <flags> section states that "--next --force bypasses safety gates."
[INDIRECT_PROMPT_INJECTION]: The --do flag allows for freeform natural language input which is then used to determine and execute system actions. This presents a risk where untrusted data could be used to manipulate the agent's behavior.
Ingestion points: The natural language string passed to the --do argument.
Boundary markers: No delimiters or safety instructions are defined to separate user intent from the agent's operational logic.
Capability inventory: The skill utilizes tools such as Bash and SlashCommand to perform its tasks.
Sanitization: The skill lacks logic to sanitize or validate the freeform intent before it is processed.
[COMMAND_EXECUTION]: The skill uses a $ARGUMENTS variable to capture and forward command-line inputs to other workflows, which may involve executing shell commands.
Evidence: The <process> block describes parsing arguments and executing specific sub-workflows.
[PRIVILEGE_ESCALATION]: The instruction to bypass safety gates serves as a mechanism to remove operational constraints, effectively granting the agent the ability to perform actions that would otherwise be blocked by its safety layer.
Evidence: The --force flag description explicitly states its purpose is to bypass safety gates.

gsd-progress