The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted output from HarmonyOS device logs and documentation searches.\n
Ingestion points: Accesses runtime data via devecocli log and devecocli docs search which could contain malicious instructions or context pollution.\n
Boundary markers: Instructions explicitly mandate the use of compact_command.py and devecocli_docs_bridge.py to process and filter output before it is included in the LLM context, rather than passing raw output directly.\n
Capability inventory: The skill is capable of executing shell commands, building and deploying applications, and performing interactive UI actions (clicks, text input, swipes) using the devecocli and hdc tools.\n
Sanitization: Log output is filtered for specific error keywords using rg, and documentation search results are processed to remove long fields and translate non-English snippets using local models.\n- [COMMAND_EXECUTION]: The skill utilizes command-line tools for application lifecycle management and environment verification.\n
Invokes the DevEco toolchain (devecocli, hvigor, ohpm, hdc) for building, linting, and device interaction.\n
Uses PowerShell for system environment checks and identifying connected device serial numbers via hdc list targets.\n- [EXTERNAL_DOWNLOADS]: The skill references external documentation hosted on the author's infrastructure for toolchain setup.\n
Provides a link to the onirodeveloper/hackyeah2026-challenge GitHub repository for CLI installation troubleshooting instructions.

ohos-app-dev