More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    AI Production Failure Reconstructor

    1

    Reconstruct production failures from logs, deploy changes, errors, and timelines to identify what happened, why, and how to reproduce it

    Free

    4 usesSecurity scanned
    AI Production Failure Reconstructor

    AI Production Failure Reconstructor

    Example session with this skill installed

    Our checkout API started returning intermittent 500 errors after yesterday's deployment.

    Here is what I have

    Error logs from the affected requests
    Deployment timestamp
    Git diff from the last release
    One successful request
    One failed request
    API response traces

    Reconstruct what happened, identify the most likely trigger, and tell me how to reproduce the failure.

    • Read your context and instructions
    • Compiled the ai production failure

    PRODUCTION FAILURE RECONSTRUCTION

    INCIDENT
    Intermittent checkout API 500 errors after deployment.

    FAILURE TIMELINE

    14:02 — New deployment completed
    14:07 — First failed checkout request
    14:09 — Successful checkout from another session
    14:13 — Repeated failures begin under the same request condition

    STRONGEST EVIDENCE
    The failing requests contain a missing payment-session field that was not present in successful requests.

    LIKELY TRIGGER
    A new checkout path introduced in the latest deployment appears to access the payment-session value before it is initialized.

    CONFIDENCE
    Medium — strong correlation exists, but the causal path requires reproduction.

    REPRODUCTION CONDITIONS
    • New checkout session
    • Payment method selected
    • Specific session state
    • New application version
    • Payment-session initialization delayed

    NEXT TEST
    Replay the failing request sequence against the new build while logging payment-session initialization.

    EXPECTED RESULT
    If the hypothesis is correct, the 500 error should occur when the payment-session value is accessed before initialization.

    REGRESSION TEST
    Add a checkout test covering the delayed-session initialization state.`

    Connects securely to your tools. The creator never sees your data.

    About this skill

    AI Production Failure Reconstructor is an evidence-first incident analysis skill for developers and engineering teams dealing with confusing production failures.

    Instead of immediately guessing a fix, it reconstructs the failure as a timeline and connects symptoms to deployments, code changes, requests, errors, environment conditions, and user actions.

    Core workflow

    Failure Timeline

    • Build a chronological sequence of events
    • Align logs, errors, deployments, requests, and user actions
    • Identify the first meaningful deviation

    Trigger Detection

    • Find conditions that may have triggered the failure
    • Compare successful vs failed scenarios
    • Separate correlation from stronger evidence

    Change Correlation

    • Compare recent deploys and code changes
    • Identify suspicious configuration or dependency changes
    • Map changes to affected behavior

    Reproduction Model

    • Convert production evidence into reproducible conditions
    • Define required inputs, environment, timing, state, and sequence
    • Handle intermittent and race-condition scenarios

    Evidence Map

    • Separate confirmed facts from hypotheses
    • Rank evidence by confidence
    • Identify missing evidence needed to confirm the theory

    Verification Plan

    • Define the smallest useful test
    • Create regression scenarios
    • Specify what evidence should appear if the hypothesis is correct

    Output

    • Failure timeline
    • Trigger hypothesis
    • Evidence map
    • Suspected change correlation
    • Reproduction conditions
    • Confidence level
    • Missing-evidence checklist
    • Verification test plan
    • Developer-ready incident handoff

    Perfect for

    • Production incidents
    • Intermittent failures
    • “Works locally” bugs
    • Post-deployment failures
    • Authentication failures
    • API errors
    • Database inconsistencies
    • Race conditions
    • Configuration-related failures
    • Performance regressions

    Important

    This skill does not pretend that a root cause is proven when evidence is incomplete. It clearly separates confirmed evidence, strong hypotheses, weak hypotheses, and unknowns.`

    Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Testing & Debugging.

    1. Cannot prove a root cause when the available evidence is incomplete

    Production-only failures may require real logs, metrics, traces, or staging reproduction
    Race conditions and distributed-system failures may require repeated testing
    Sensitive production data should be redacted before analysis
    Final deployment and production changes should remain under human review`

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before buying a skill.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    4 uses

    Times used to date

    Free to use

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 3 months ago

    • Free with an Agensi account

    Permissions

    Read Files

    File Scopes

    **/*.log
    **/*.txt
    **/*.json
    **/*.yaml
    **/*.yml
    **/*.md
    **/*.diff
    **/*.patch
    Listed3 months ago

    What's inside

    Frequently Asked Questions

    This skill takes your raw logs, deployment data, and error reports to build an evidence-based timeline. It prevents 'guess-and-check' debugging by identifying the exact moment of deviation and mapping it to specific code or configuration changes.

    The Reconstructor works with any LLM capable of processing structured data, but it performs best with models that have large context windows (like Claude 3.5 Sonnet or GPT-4o) to ensure it can analyze high volumes of log data simultaneously.

    No complex installation is required; you simply provide the skill with your incident data (logs, error traces, and deployment timestamps). It is designed to be 'plug-and-play' within your existing AI agent workspace or CLI environment.

    Unlike generic tools, this skill uses a 'Missing-Evidence Checklist' to highlight what you don't know yet. It explicitly ranks hypotheses by confidence and refuses to claim a root cause is found until the evidence map is complete.

    Your purchase includes the core reconstruction logic, a standardized incident handoff template for developers, and documentation on how to format your evidence for the highest accuracy analysis.

    New in Testing & Debugging

    Free