The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes output from external command-line tools, which creates a surface for indirect instructions to influence agent behavior if the output is manipulated.
Ingestion points: The agent reads and classifies output from joule list/status, terraform output, and btp security/role-collection command results (SKILL.md).
Boundary markers: The skill does not define specific delimiters or instructions to ignore potential commands embedded in tool outputs.
Capability inventory: The agent has the ability to execute shell commands (terraform apply, local scripts) and modify .tfvars configuration files.
Sanitization: There is no evidence of sanitization or validation logic applied to the tool outputs before they are used to determine next steps.
[COMMAND_EXECUTION]: The skill is designed to execute local shell scripts provided within the project infrastructure.
Evidence: The instructions call for executing ./scripts/joule-login.sh and infra/btp/scripts/check-plan.sh. While these are local project scripts, they represent an execution capability that relies on the integrity of the local file system.

sap-repair-joule-access