The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill determines the project's technology stack by ingesting metadata from several common project files, creating a surface for indirect prompt injection where malicious instructions could be embedded in external data.
Ingestion points: Project files such as package.json, pubspec.yaml, *.xcodeproj, Package.swift, build.gradle.kts, and build.gradle (specified in SKILL.md).
Boundary markers: None observed in the provided instructions for handling these files.
Capability inventory: The skill includes extensive capabilities for file manipulation, network operation guidelines (Retrofit/Dio), and shell-based automation for Android devices/emulators.
Sanitization: No specific sanitization or validation of the ingested metadata is described before it influences the agent's behavior.
[COMMAND_EXECUTION]: The references/android-rules/android-emulator-skill.md file defines a workflow utilizing local shell and Python scripts for Android device automation.
Evidence: Mentions scripts such as scripts/emu_health_check.sh, scripts/app_launcher.py, scripts/screen_mapper.py, and scripts/navigator.py to be executed via bash or python for emulator management and semantic UI navigation.
[EXTERNAL_DOWNLOADS]: The skill references several external resources and best practice repositories from GitHub.
Evidence: The skill incorporates rule sets and documentation from repositories including vercel-labs/agent-skills, kevmoo/dash_skills, new-silvermoon/awesome-android-agent-skills, and AvdLee/SwiftUI-Agent-Skill. These are utilized as static reference material for the agent's mobile development logic.

mobile-developer