Operational Metrics Guided Review

    1

    Compare approved operational outcome and guardrail rates with weighted denominators, missingness, cohort definitions and fixed windows. Separate descriptive signals from incomplete evidence.

    Free

    Security scanned

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    Operational Metrics Guided Review

    Example session with this skill installed

    Complete the task in the included fictional example, using only its supplied evidence and taking no external actions.

    • Read your context and instructions
    • Compiled the operational metrics guided

    Final actual agent review: revised metrics package

    Status: human review required; semantic support/source authenticity UNVERIFIED; causal attribution NOT_ESTABLISHED.

    Definitions and scope

    EU fixed at intake, v1, cancellations excluded, invoices counted once in their metric-specific daily bucket. Late measures observed eligible invoices finally closed in bucket with known intake/closure timestamps; numerator strictly >24h UTC wall-clock original intake to final closure including reopenings; missing captures unknown/unclassifiable timestamps. Approval return uses final-decision-date buckets after reopenings, complete classifiable history denominator, at least one explicit return-for-correction numerator, final-decision history cutoff and missing history counts. These populations differ, so denominators need not match. Do not pool the two metrics.

    Baseline [2026-10-01,2026-10-03), current [2026-10-03,2026-10-05), UTC; 24h reporting lag. As-of Oct 6 00:00 is at the latest required cutoff, so all supplied days are closed and past the required lag as the example asserts. Daily records asserted unique, finalized, census-complete. This does not authenticate source/approval or prove completeness. Scope is two metrics, one cohort, two days per window, eight expected cells, seven supplied; within manual limits.

    Gates and arithmetic

    Late B1+B2: (1+9)/(10+90)=10/100=0.10. Missing 0, eligible 100, coverage 100/100=1. Current C1+C2: (5+9)/(10+90)=14/100=0.14, missing 0, eligible 100, coverage 1. Each daily denominator >=10 and coverage 1; compatible v1 and no evident source-wide invalidity in supplied rows.

    Change is +4/100=+0.04, below minimum deterioration 5/100. Weighted current threshold test is strict 14/100>15/100, false. Daily current rates C1=5/10=0.50 and C2=9/90=0.10; only C1 breaches 0.15, not a two-day run. Revised guide explicitly excludes isolated daily breach from independent attention. Late priority none under these configured triggers. This is neither health nor efficacy assurance.

    Guardrail G1+G2 baseline 2/100=0.02, missing 0, eligible 100, coverage 1. Current G3 alone 1/50=0.02, missing 0 and eligible 50 in that supplied row. G4 absent: partial totals only, full-window denominator/missingness/coverage unavailable. Do not use observed-row coverage 1 as window coverage, fill G4, or compare complete windows. Guardrail priority review_data independently; no overall favorable conclusion.

    Interpretation and next checks

    Reconcile census/approval assertions and obtain G4. C1's 50% is a descriptive excursion, not configured attention, recurrence or cause. Review case mix, seasonality, selection and measurement/operational change before interpretation. Aggregate IDs are not case IDs; no automatic case investigation, expanded window or downstream skill invocation. No thresholds changed, approval hashes invented, source action, scheduling or notification performed.

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

    About this skill

    What you get

    Compare approved operational outcome and guardrail rates with weighted denominators, missingness, cohort definitions and fixed windows. Separate descriptive signals from incomplete evidence.

    Includes a tailored evidence/decision guide, report template and bounded fictional practice input. The host agent reviews source-qualified observations and uncertainty, then prepares a provisional result for a human decision.

    Requirements and scope

    Use an Agensi-connected agent that can read delivered text. No Python, CLI or installation is required. Real reviews need sanitized supplied evidence or a separately authorized existing read integration.

    Use appropriately sanitized supplied evidence by default. Existing read tools may be used only within separately authorized access; this skill grants no access or sharing permission.

    Limitations

    This is an installation-free, guidance-only adaptation of the original public MIT-licensed skill. It does not include the original deterministic runtime, automatic redaction, schema/quote validation, integrity hashing or enforced budgets. It does not authenticate approvals, prove causes, certify compliance or carry out recommendations.

    Source and license

    Source: https://github.com/evgenytkachenko/operational-quality-skills/tree/d330c2956c56c810e6cb8fceb95fe5b456b1b76a/skills/operational-quality-monitor

    Original author: Evgeny Tkachenko (evgenytkachenko). Copyright (c) 2026 evgenytkachenko. MIT license included in the package.

    Verification

    Historical demonstration: a local text-loaded exercise using the included fictional input. This was not a behavioral test using Agensi-delivered instructions or validation of the original CLI. The listing is live, but behavior using Agensi-delivered instructions and independent buyer access have not been tested.

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

    Guidance-only; semantic support UNVERIFIED; human review required.
    No original CLI, deterministic schema/quote/arithmetic checks, automatic redaction, verified hashes or machine-enforced limits.
    No source/approval authentication, causal proof, compliance certification or demonstrated operational effectiveness.
    No system changes, implementation, schedules, notifications, external communications or new permission grants.
    This listing is live. Behavior using Agensi-delivered instructions has not been tested. Historical local text-loaded exercises do not establish reliability across other agents or operational effectiveness.

    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

    Fresh listing

    Recently published to Agensi

    Free to use

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 10 days ago

    • Free with an Agensi account

    Software
    Python

    Permissions

    Read Files

    Read only the delivered package text and sanitized supplied evidence, or an independently authorized existing read source. No Python, shell, installation, additional permissions or integrations, system changes or external actions.

    Listed10 days ago

    What's inside

    Frequently Asked Questions

    Compare approved operational outcome and guardrail rates with weighted denominators, missingness, cohort definitions and fixed windows. Separate descriptive signals from incomplete evidence. All findings are provisional and require human review.

    Use an Agensi-connected agent that can read the delivered text. The example uses supplied fictional evidence. Real reviews need sanitized supplied evidence or a separately authorized existing read integration. No Python, CLI, installation or additional integration is required for the included example.

    The free package includes SKILL.md, a tailored guide, a report template, a fictional example, README and the original MIT license. Original author: Evgeny Tkachenko (evgenytkachenko). It excludes the original runtime, automatic redaction, deterministic checks and enforced budgets. The displayed example is a historical local text-loaded exercise. Behavior using Agensi-delivered instructions has not been tested. Neither the local exercise nor listing approval proves causal validity or operational effectiveness.

    New in Business & Operations

    Free