The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill documentation references the use of an external MCP tool, @tencent-rtc/trtc-push-mcp, to handle the logic of the integration workflow. This package is hosted under the official Tencent RTC scope on npm, aligning with the skill's authorship.
[COMMAND_EXECUTION]: The integration instructions guide the user to execute platform-standard build and dependency commands, such as ./gradlew for Android and pod install for iOS. These are necessary and expected for the primary purpose of the skill.
[INDIRECT_PROMPT_INJECTION]: The skill accepts user-provided prompts which are forwarded to the external workflow engine. While this represents a data ingestion surface, the risk is mitigated by the narrow domain of TIMPush integration and the use of a vendor-provided logic engine.
Ingestion points: user_prompt field in SKILL.md passed to the get_workflow_state tool.
Boundary markers: Not explicitly defined in the provided instruction text.
Capability inventory: Significant; the workflow engine can suggest file modifications and command executions to facilitate SDK setup.
Sanitization: Not specified, relying on the underlying LLM's safety filters.
[SAFE]: The skill explicitly warns against hardcoding secrets like SDKAppID or AppKey. It provides templates and instructions to use local.properties (Android) and gitignored source files (iOS) to protect credentials, which is an industry-standard security practice.

trtc-push