The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user or agent to use the submit-expo-feedback utility via npx to report issues. This triggers a remote download and execution of the package from the npm registry. This is a vendor-owned feedback tool intended for skill improvement. Evidence: SKILL.md (Line 49).
[EXTERNAL_DOWNLOADS]: Setup instructions include the installation of the expo-observe library using npx expo install. This is a standard procedure for adding native modules to Expo projects. Evidence: references/setup.md (Line 38).
[INDIRECT_PROMPT_INJECTION]: The skill exposes the agent to indirect prompt injection risks by instructing it to query and interpret performance data and custom events via the EAS CLI. If an attacker controls data emitted by an app (such as custom event names or properties), they could embed malicious instructions that the agent might obey when processing the CLI output. Ingestion points: CLI output from eas observe:metrics, eas observe:events, and eas observe:routes as described in references/queries.md. Boundary markers: Absent; instructions direct the agent to read and interpret the output directly without specific delimitation or ignore-directives. Capability inventory: Shell execution for CLI queries (eas) and feedback submission (npx). Sanitization: No sanitization or validation of the ingested CLI data is specified in the skill instructions.

eas-observe