{"id":73128,"plugin_id":"plugin_asdk_app_6acabb9e80c48191bfe0cde0cab66df1","kind":"skill","collection_source":"plugin_package","comparison_source":null,"observed_at":"2026-10-31T01:18:42.005Z","digest":"3f8067b09aa99d990e5e1e5140af5d0ae9ed4f0d1493befc924cdcadd3b94d63","against":null,"payload":{"description":"Diagnose problems on a paired computer, server, or VM. Use when the user wants to inspect a failing app, process, local logs, configuration, or machine state and troubleshoot or fix the issue.","included_files":[{"relative_path":"agents/openai.yaml","size_in_bytes":647},{"relative_path":"assets/icon.png","size_in_bytes":3037}],"name":"remote-device-troubleshooting","skill_md_contents":"---\r\nname: remote-device-troubleshooting\r\ndescription: \"Diagnose problems on a paired computer, server, or VM. Use when the user wants to inspect a failing app, process, local logs, configuration, or machine state and troubleshoot or fix the issue.\"\r\n---\r\n\r\n# Remote device troubleshooting\r\n\r\nUse this skill when the user wants you to inspect, diagnose, or help fix a problem on a computer, server, or VM they own or administer.\r\n\r\n1. Resolve the target machine.\r\n   - If the target device is already unambiguous from the conversation, keep using it.\r\n   - Otherwise call list_devices.\r\n   - Ask the user only when multiple plausible paired devices remain and choosing the wrong one could matter.\r\n\r\n2. Start with the least invasive evidence that can answer the question.\r\n   - Use ping, get_config, list_processes, get_file_info, list_directory, read_file, read_multiple_files, or search tools as appropriate.\r\n   - Prefer evidence from the failing process, relevant logs, config, and nearby files over broad system sweeps.\r\n   - Do not invent a root cause when the available evidence is inconclusive.\r\n\r\n3. Use commands only when they materially improve diagnosis or execute a user-requested fix.\r\n   - Use start_process for targeted diagnostics, builds, CLIs, or other commands that the user wants run.\r\n   - Read long-running output with read_process_output; use list_sessions when a process may already be running.\r\n   - Treat tool errors and safety-policy rejections as real outcomes. Do not weaken RelayDesk's local policy to force a command through.\r\n\r\n4. When the user asks to fix the problem, make the smallest justified change.\r\n   - Explain or infer the intended change from the user's goal.\r\n   - Use file, configuration, or process mutation tools only when necessary.\r\n   - Re-check the affected process, file, or command after the change so the final answer distinguishes \"changed\" from \"verified fixed.\"\r\n\r\n5. Respect RelayDesk's security boundary.\r\n   - Never use RelayDesk to obtain passwords, API keys, auth tokens, MFA or OTP codes, passphrases, private keys, payment-card data, protected health information, or government identifiers.\r\n   - Do not ask to change blockedCommands or allowedDirectories remotely.\r\n   - If authentication is required, tell the user to complete it locally on the paired device.\r\n\r\n6. Report a compact evidence-based result:\r\n   - target device,\r\n   - what was checked,\r\n   - evidence found,\r\n   - action taken, if any,\r\n   - verification result,\r\n   - remaining uncertainty or next safe step.\r\n\r\nDo not invoke RelayDesk merely because the user asks a technical question. Use it when the task genuinely requires access to one of their paired machines.\r\n"},"changes":[],"summary":"First saved snapshot. No earlier version is available for comparison.","summary_kind":"deterministic","summary_metadata":{}}