About this skill

Problem: Operations reviews mix results, causes, and actions

Operations retrospectives often stall because “the numbers look right, but the explanations are scattered.” Target definitions, baselines, process records, and anomaly logs live in different places; facts and assumptions get written as conclusions; and the final output becomes slogans instead of decisions about what to keep, fix, or stop. This skill treats the review as an executable artifact: restore the process with consistently scoped data, then separate facts, deviations, causal hypotheses, and action priorities.

How it works: input gates to validated output

It follows a fixed order: read references/methodology.md, references/execution-rules.md, references/quality-boundaries.md, references/output-contract.json, and assets/output-template.md, selecting only the method cards directly relevant to the current request. When inputs are incomplete, it distinguishes confirmed facts, conservative assumptions, items to verify, and evidence dates instead of guessing. If routing is unclear, it asks one key question.

A typical deliverable includes:
- Target comparison: original target, success metric, baseline, and plan version
- Evidence gaps: data definitions, result data, process records, timeline, and anomaly logs
- Deviation analysis: material deviations, impact scope, and counter-evidence
- Action recommendations: keep, fix, stop, and retest plan

After generation, it runs scripts/validate_output.py --check-output to check structure, specificity, and known risks. Passing validation does not mean the business owner has confirmed fit.

Boundaries

It fits team retrospectives with basic targets, data, process records, and decision logs. It does not invent cases, metrics, or industry benchmarks. It avoids copying article text as a solution and does not expose collector identity, raw links, API keys, or source material.

Screenshots

Use Cases

Before an ops weekly, compile targets, baselines, results, and anomaly logs into a fact timeline.
During channel reviews, separate confirmed facts, assumptions, and verification items to avoid over-attribution.
Before launch, use target checks and deviation impact to define keep, fix, stop, and retest actions.
When metric scope conflicts, produce evidence gaps and evidence dates for owner confirmation.

Best For

Operations owners who need to turn targets, data, and anomaly logs into a deliverable review.
Growth analysts who need evidence gaps and verification items when metric scope is inconsistent.
Product leads who need to judge deviation impact and choose keep, fix, or stop actions.
Data engineers who need a validated retrospective structure with evidence dates retained.