The Agent Skills Directory

[PROMPT_INJECTION]: The skill metadata claims the author is 'Google LLC', which is deceptive as it contradicts the actual developer context ('talissonvitorino'). This impersonation could lead a user or agent to trust the skill's instructions based on a false identity.
[DATA_EXFILTRATION]: The skill directs the agent to access and inspect files within the user's home directory at ~/.gradle/caches/modules-2/files-2.1/ to verify dependency metadata. Accessing file paths outside the project root poses a privacy and data exposure risk.
[COMMAND_EXECUTION]: The skill requires the execution of shell commands, including ./gradlew help and ./gradlew build --dry-run, to verify the success of the migration process.
[EXTERNAL_DOWNLOADS]: The skill references and encourages the use of the gradle-recipes repository from a well-known service (GitHub) to obtain migration templates and code examples.
[PROMPT_INJECTION]: The skill processes untrusted project configuration data (such as build scripts and gradle.properties) while maintaining the capability to execute shell commands, creating a surface for indirect prompt injection attacks.
Ingestion points: Project-level build scripts (.gradle, .gradle.kts) and gradle.properties files.
Boundary markers: None identified.
Capability inventory: Shell execution via the Gradle wrapper (./gradlew).
Sanitization: No explicit sanitization or validation of the content read from the project files is defined.

agp-9-upgrade