amio/dot-agents/diagnose-live-app-behavior
Claude Code and agent skill by amio/dot-agents
- community
Checking that the SKILL.md file is still in the repository.
Diagnose and fix real-world macOS GUI issues by launching an app with actual data, letting the user reproduce the problem once, and observing it live. Only use when the user explicitly requests this live-diagnosis flow.
Install
npx vskill@latest install amio/dot-agents/diagnose-live-app-behaviorSKILL.md
Diagnose Live App Behavior
Capture the real scenario
- Launch or identify the real-data production app; preserve its data. Record the executable/build, PID, relevant environment, exact action, expected result, and visible symptom so evidence belongs to the process the user actually operates.
- Prepare suitable capture tools before asking the user to reproduce once. Mark a short start/end interval and correlate the action with timestamps, relevant object identity, and thread/queue context where available. Capture the symptom and enough surrounding execution to locate its cause.
- When Xcode debugging is available, use Debug → Attach to Process for an existing process, or the scheme's Run → Info → Debug executable option for a launch. Verify attachment and symbol availability. If the production build cannot be attached, collect available logs, samples, or reports first; identify any later debug-build evidence as a different execution environment.
Choose evidence by symptom
Use the smallest relevant set of tools, not every diagnostic on every run. Check tool and target support before relying on a feature.
| Symptom | Evidence and next step | |
|---|---|---|
| Incorrect state or result | Break just before the suspected transition, inspect inputs, then step to the first unexpected value. If inputs are already wrong, move upstream to their producer. Use conditions or ignore counts to isolate the relevant occurrence. | |
| Intermittent or concurrent behavior | Prefer existing logs or bounded tracing. When using breakpoint actions, enable Automatically continue after evaluating actions. Prefer v for variable inspection; po dynamically evaluates expressions and costs more. Log messages can include context and hit counts. Even auto-continuing breakpoints perturb timing; disappearance under a debugger is not proof of a fix. | |
| Crash, exception, or runtime issue | Preserve the diagnostic message and stack. A stop at main or AppDelegate may be downstream of an Objective-C exception or runtime issue, including in Swift apps. Select an Objective-C Exception, Swift Error, or Runtime Issue breakpoint based on evidence to catch the origin. A Swift Error breakpoint catches throws, not every Swift trap. Runtime Issue breakpoints need the corresponding supported checker/sanitizer enabled; they do not detect all faults by themselves. | |
| Hitches or unexpected resource use | Compare CPU, memory, Disk I/O, and Network I/O gauges across idle, action, and recovery. Investigate persistent idle activity or CPU spikes aligned with hitches using targeted profiling or samples. Compare I/O with expected payload and operation counts; trace repeated reads, downloads, retries, or writes to lifecycle and cache ownership. Gauges identify where to investigate, not the root cause. | |
| Memory keeps growing | Capture the memory graph after the relevant lifecycle transition; it pauses execution. Inspect unexpected surviving instances and follow strong incoming references. Distinguish a retain cycle from abandoned objects that remain reachable but are no longer needed. Change a reference to weak only when ownership permits it; otherwise remove the dependency or release references at their proper lifecycle boundary. Growth alone does not prove a leak. | |
| Wrong appearance, position, or size | Capture Debug View Hierarchy while the faulty view is present; inspect the actual layers and selected view. Follow frame calculations for frame-based layout, active constraints for Auto Layout, and resolved size/placement for SwiftUI. Use supported Environment Overrides for the relevant appearance, text-size, or accessibility setting. Apply platform-appropriate tools; do not copy UIKit-only symbols into an AppKit diagnosis. |
Establish cause and fix the owner
Correlate runtime evidence with the user's marked symptom. Use static review to form hypotheses, then identify the first incorrect state transition, reference relationship, or repeated operation that explains the observation. Separate observed facts from inferred causes and state what evidence is missing if the capture is inconclusive.
Fix the authoritative owner boundary and avoid speculative local patches. If more evidence is necessary, first use captured artifacts or agent-controlled reproduction; do not silently turn the one-reproduction workflow into repeated user requests.
Verify and hand off
Run proportionate automated or agent-controlled checks that address the observed failure and do not require another user reproduction. Compare the relevant before/after signal when feasible, including state, resource recovery, object lifetime, or layout. If the change fails the check, revisit the diagnosis. Passing unrelated tests does not establish that the live symptom is resolved.
Document non-obvious owner rules beside the governing code. Remove only diagnostics introduced for this investigation and restore changed debugger settings or environment overrides. Report the evidence, cause, fix, checks, and remaining uncertainty. Offer same-scenario live validation only as an optional follow-up; explicitly say when the original live scenario has not been revalidated.
Apple references
The diagnostic choices above adapt these Apple documents to this skill's macOS, real-data, single-user-reproduction workflow. Consult them when detailed debugger setup is needed:
Security
- Security scanPassed, 90 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingskills/diagnose-live-app-behavior/SKILL.md
Popularity
- 7-day trend score
- 0
Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/amio/dot-agents/diagnose-live-app-behavior)