The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to query and process configuration data from external services like Firebase and Supabase, which could contain maliciously crafted metadata intended to influence the agent's behavior.
Ingestion points: The agent is directed to use firebase and supabase MCP tools to fetch collections, providers, tables, and schemas in the workflow section of SKILL.md.
Boundary markers: There are no explicit instructions or delimiters provided to safeguard the agent from instructions potentially embedded in the data returned by these tools.
Capability inventory: The skill enables the agent to run local build and test commands like flutter test and dart run build_runner.
Sanitization: No validation or filtering is recommended for the metadata retrieved from external backends prior to its use in generating application code.
[COMMAND_EXECUTION]: The skill requires the agent to execute multiple local CLI commands for project management and validation.
Evidence: The integration workflow involves running dart format ., flutter analyze, flutter test, and dart run build_runner build --delete-conflicting-outputs as specified in the JSON Serialization and Workflow sections.

flutter-api-integration