The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the local repository, including AGENTS.md, .scenes.json, and Gradle configuration files, to identify build commands, package names, and project structures.
Ingestion points: The agent is instructed to read AGENTS.md, build files, and manifest files at the start of the workflow.
Boundary markers: The skill includes warnings such as "Treat AGENTS.md as the local navigation source, not as a substitute for verification" and "Verify current behavior through the normal development loop; an earlier runnable handoff does not establish that the app still builds or launches."
Capability inventory: The skill utilizes Bash (for gradlew, adb, git, and pico-cli), Read, Write, and Edit permissions.
Sanitization: No explicit sanitization or validation logic for the content of the project files is described beyond the recommendation to verify behavior manually.
[COMMAND_EXECUTION]: The workflow relies on shell command execution via Bash for critical development tasks.
Evidence: The skill executes ./gradlew assembleDebug for builds, pico-cli for bundle verification and performance analysis, and adb (delegated via spatial-emulator-usage) for device management, installations, and log collection.

spatial-app-dev-workflow