From lightrun-ai
Instruments running services with Lightrun to answer live production questions about variable values, execution durations, hit counts, and value distributions.
How this skill is triggered — by the user, by Claude, or both
Slash command
/lightrun-ai:lightrun-ask-prodThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Query live production runtime to answer questions about system behavior using Lightrun's observability tools.
Query live production runtime to answer questions about system behavior using Lightrun's observability tools.
Read the user's question and identify:
Select the Lightrun capability that best fits the question. More than one may be needed.
| If the question asks... | Use... |
|---|---|
| What is the current value of X? | Snapshot expression capture (snapshot_create, snapshot_status, and snapshot_get_values) |
| How long does operation X take? | Execution duration (get_runtime_execution_duration) |
| How often does line X run? | Execution count (get_runtime_execution_count) |
| What range of values does X take over time? | Distribution / numeric metric (get_runtime_numeric_metric) |
Before calling any tool, identify the specific file path and line number that will yield the needed information.
Useful signals by question type:
If the code location is ambiguous, ask the user to clarify before proceeding. Do not guess.
Call get_runtime_sources to retrieve all agent pools, each with their agents, tags, and custom sources in one response.
Apply this selection logic:
Do not present a single-instance result as a global production answer. When full coverage is not available, describe the observed scope explicitly.
Call the appropriate Lightrun tool with:
Use runtime actions for evidence collection:
Interpret the returned data in the context of the user's question:
| Situation | Action |
|---|---|
| No agent pools found | Inform the user; ask them to verify the service is running and connected to Lightrun |
| Multiple plausible agent pools | Present the list to the user and ask which to use |
| Multiple plausible agents within a pool | Use the agents, tags, and custom sources returned by get_runtime_sources to narrow down; if still ambiguous, ask the user |
| A single source cannot answer a fleet-level question | Target a tag, custom source, or multiple agents that match the requested scope; if full coverage is unavailable, frame the answer in terms of selected sources |
| Tool returns no data | Keep the action active for a longer observation window; if still empty, consider whether the code path is being actively exercised |
| Longer observation window needed | State the observation window, action ID, selected source scope, signal being collected, and condition for checking results later |
| Line cannot be instrumented | Try an adjacent line; if still unavailable, explain the limitation to the user |
| Code location unclear | Ask the user before proceeding — do not guess |
cacheSize in the selected production instances.POST /api/v1/users.varX is true?calculateDiscount runs for customer X?npx claudepluginhub lightrun-platform/lightrun-ai --plugin lightrun-ask-prodPlaces live log probes on running services to capture runtime argument/variable values without redeploying. Useful for inspecting live data from production methods.
Guides runtime debugging in live environments using Lightrun MCP, with problem framing, hypothesis ranking, evidence capture, diagnosis confidence, and blocker handling.
Tracks response times across API endpoints, database queries, service calls, and more with P50/P95/P99 percentiles and SLO monitoring to identify bottlenecks.