More screenshots
Works with the AI tools you already use
Claude Code
Cursor
Codex CLI
GitHub Copilot
Gemini CLI+20 moreClaude CodeCursorCodex CLIGitHub CopilotGemini CLI
OpenClaw
Manus
Grok+17 agentsObservability Reproduction Designer
Turn intermittent incidents into safe telemetry and falsifiable tests.
$19.99
Observability Reproduction Designer
Example session with this skill installed
Design a reproduction plan for intermittent checkout timeouts after payment authorization. We have UTC millisecond timestamps, gateway status and latency, region, payment type, cart-size bucket, release ID, and a generic timeout error. Gateway events cannot be joined to worker attempts. The worker retry ordinal and order-state transition are missing. Raw customer and payment fields must never be logged. Create competing hypotheses, the minimum telemetry delta, a controlled protocol, and stop rules.
- Read your context and instructions
- Compiled the observability reproduction designer
Verdict — INSTRUMENTATION-REQUIRED
The symptom is precise enough to form competing hypotheses, but current evidence cannot join a late gateway response to a specific worker attempt or order-state transition. One bounded event is required before a responsible reproduction attempt.
Symptom contract
Intermittent checkout timeout observed after payment authorization. Known-good comparator: successful checkouts with the same region, payment type, cart-size bucket, and release ID. Success criterion for reproduction: controlled execution reaches authorization, produces overlapping or delayed worker behavior, and ends in the same timeout category and terminal order state.
Evidence map
| Domain | Status | Evidence |
|---|---|---|
| Correlation | gap | gateway and worker identifiers cannot be joined |
| Time | available | UTC millisecond timestamps |
| Cohort | gap | region available; tenant grouping absent |
| Input | available | bounded cart size and payment type |
| Application state | gap | order transition not emitted |
| Dependency state | available | gateway status and latency |
| Retry | gap | attempt ordinal missing |
| Error | available | timeout category |
| Version | available | release ID |
| Privacy | available | raw payment data excluded |
Competing hypotheses
H1 — A worker retry overlaps a late gateway response. Support: failures follow authorization. Contradiction: some samples show only one worker log. Falsifier: correlated failures contain no overlapping attempt ordinals.
H2 — One region experiences dependency latency spikes. Support: weak regional clustering. Contradiction: the failure sample is small. Falsifier: matched success and failure cohorts show equivalent gateway-latency distributions.
Minimum instrumentation
Emit checkout_transition immediately before and after the gateway boundary with hashed_correlation_id, bounded state_from and state_to, attempt_ordinal, region, and release_id. Capture 100% of failures and a 1% success comparator for 14 days. Exclude raw customer, cart, token, and payment fields. Bound all dimensions to enumerations or an approved hash. Remove the event after the hypothesis verdict and retention review.
Controlled reproduction
Freeze a gateway stub at selected late-response intervals, replay one bounded cart fixture, run attempts one and two with a controlled overlap, then compare event order and terminal state against matched success cases.
Stop rules
Stop immediately if any unredacted sensitive field appears, if cardinality exceeds the agreed cap, if the test affects a real customer, or when both hypotheses are falsified or one is reproduced three controlled times.
Validator
The plan contains all ten evidence domains, two falsifiable hypotheses, a complete instrumentation contract, protocol, and stop rules and passes the bundled structural validator. Nothing was deployed or executed.
Connects securely to your tools. The creator never sees your data.
What you get
About this skill
More logs do not automatically create better evidence
Broad debug logging can add noise, privacy exposure, high cardinality, and cost without separating competing causes. Elusive incidents need a precise symptom, a known-good comparator, falsifiable hypotheses, and the smallest safe observation that changes the decision.
Observability Reproduction Designer maps correlation, time, cohort, input, application state, dependency state, retries, errors, version, and privacy. It defines temporary instrumentation with fields, location, sampling, retention, redaction, cardinality limit, and removal condition, then writes a controlled reproduction protocol and stop rules.
Built for SRE, platform, backend, and incident-response teams handling intermittent, tenant-specific, timing-sensitive, or dependency-related failures. It profiles the existing trace, log, metric, privacy, test-harness, retry, and release conventions before recommending any new signal.
The package includes full Skill source, evidence-map method, worked checkout incident, deterministic validator, mutation tests, and a six-page Buyer Guide PDF.
Boundaries
It does not access production, deploy telemetry, run load, collect private customer data, or claim a root cause. A complete plan is not a reproduced incident.
How to install
Works the same in every agent - Claude, Cursor, Codex, Copilot and 20+ more.
- 1
Download the ZIP
Free skills download straight away. Paid skills unlock right after purchase.
- 2
Unzip into your skills folder
Every agent reads skills from one folder on your machine. Drop the unzipped folder in there.
- 3
Ask your agent to use it
Restart the agent if it was already running. It picks the skill up automatically - no config needed.
Skills folder by agent
Click the path to copy it. Create the folder if it does not exist yet.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill passes our trust checks below.
Security scanned
Passed our 8-point scan before listing
Fresh listing
Recently published to Agensi
30-day refund
Not a fit? Get your money back
Trust & safety
Security scanned
Verified clean 16 days ago
- Passed all security checks, Safe to install