The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local vendor-provided command-line tools to manage the environment discovery and manifest generation workflow.
Evidence: Executes ./cookbookctl for pilot status, account discovery, manifest validation, and local rendering of configuration files.
Evidence: Executes btp login --sso to facilitate secure user authentication via the official SAP BTP CLI when a session is not active.
Evidence: Runs scripts/region-preflight/region-preflight.sh to verify service availability in specific BTP regions before configuration.
[INDIRECT_PROMPT_INJECTION]: The skill ingests configuration data from local template files to scaffold agent manifests, presenting a surface for indirect prompt injection that is mitigated by local storage and human review.
Ingestion points: Reads from templates/hr-leave.yaml, templates/inventory.yaml, templates/procurement.yaml, and templates/sales-orders.yaml at runtime.
Boundary markers: Instructions mandate that the agent must present a summary for user confirmation before any pilot.yaml manifest is written to the filesystem.
Capability inventory: Shell command execution via btp and cookbookctl, and writing to local files (pilot.yaml, generated.tfvars.json).
Sanitization: The agent is instructed to derive kebab-case identifiers and strictly follow user-confirmed architectural decisions.
[SAFE]: The skill contains explicit instructions to prevent the handling of sensitive credentials.
Instructions explicitly state: "Never request backend secrets in chat" and "never accept passwords, tokens, service keys, certificates, or API keys in chat or pilot.yaml."

sap-setup-btp-agent