Files
documentation/docs/strategy/skills-maintenance.mdx
jarvis-at-skic e22d3a5c13
Build and publish Eldorado documentation / build (pull_request) Successful in 52s
docs: separate workflow effectiveness from skill health
2026-10-08 11:05:00 +00:00

65 lines
5.3 KiB
Plaintext
Raw Permalink Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
---
id: skills-maintenance
title: Skills maintenance
description: Review, drift, deprecation, rollback and incident handling.
hide_table_of_contents: true
---
import ArchifyDiagram from '@site/src/components/ArchifyDiagram';
<div className="strategy-content">
**TO-BE · Target operating model — not installed controls or company results.**
## What happens when a skill drifts or causes harm?
Monitor both **Workflow effectiveness** (comparable outcomes, safety, cost and human experience) and **Skills effectiveness & health** (selection, supported use and substantive review) throughout a revision's life. Use the evidence to improve, retain or retire skills; adoption or current review dates alone do not prove better work. Unsafe behavior triggers containment; a correction returns through independent tests and approval, never automatic reinstatement.
<ArchifyDiagram name="strategy-maintenance" title="Monitor, triage and test corrections; promote safely or contain and roll back" story />
- **Named ownership:** Data &amp; AI coordinates catalog health with the owning teams.
- **Early triggers:** incidents and client, model or dependency changes override the calendar.
- **Recovery is verified:** rollback restores a known-good bundle; it cannot undo disclosure or past effects.
<details>
<summary>Review responsibilities, incident handling and proposed review intervals</summary>
**TO-BE · Target operating model.** Introduction depends on the [roadmap gates](./roadmap.mdx); these pages do not describe installed controls or measured company results.
Every approved skill has an accountable owning team, independent reviewer, review due date, supported client matrix and rollback bundle. Review both usefulness and safety, not just metadata freshness. Proposed maximum review intervals are recorded by risk tier; incidents, dependency changes, model/client drift or repeated failures trigger earlier review.
Data & AI reviews catalog health with the owning teams. Quality Engineering samples accepted outputs and adds confirmed failures to regression sets without exposing sensitive content. Platform & IT watches unsupported, deprecated and revoked revisions and tests distribution recovery. Architecture & Security reviews emerging misuse and exception expiry. Remove redundant or unused entries deliberately, after examining whether poor discoverability or fit explains non-use.
A reported unsafe behavior opens an incident with artifact, policy and client versions. Architecture & Security coordinates containment; Platform & IT revokes authority, stops scheduling and quarantines affected revisions; the owning department handles customer or business consequences; Quality Engineering reproduces the failure; Data & AI coordinates catalog correction and trend analysis. Preserve minimal restricted evidence and involve People & Administration Squad for employee-data concerns and Finance & Accounting Squad for financial effects. Restart requires a reviewed fix, regression evidence, verified rollback/revocation and accountable human authorization. Closing the incident does not erase the historical failure from measurement.
## Proposed maximum review intervals
Owner ratification is required before adoption. Review earlier after incidents or material dependency changes.
| Tier | Maximum days between substantive reviews |
| --- | --- |
| R1 | 90 |
| R2 | 60 |
| R3 | 30 |
</details>
<details>
<summary>Ongoing reporting and comparison</summary>
## Keep selected evidence useful
Continue the measures agreed for the workflow in [Baseline](./baseline.mdx) and applied during [Adoption](./skills-adoption.mdx). Revisit relevance when scope changes; the [registry](./measurement/index.mdx) is a source of choices, not a requirement to collect every metric. Preserve current-workflow comparisons separately from named skill/version evaluations. No applicable skill population is N/A, not zero; missing evidence for existing revisions is unknown and must not improve reported health.
If substantive-review coverage is selected, join each active approved catalog revision to its owner, substantive review evidence and due date. Apply [review coverage](./measurement/skills-within-substantive-review-window.mdx): revisions with all three current / all active approved revisions × 100. The proposed Data & AI review is monthly and at each release, subject to agreement. Report overdue/ownerless revisions and deprecated or revoked versions still in use—not just a percentage.
At each agreed report, compare like workflows with their recorded starting position. Publish period, counts, exclusions, source coverage and changes in calculation or scope. Zero denominators mean “not available”; unresolved gaps remain visible. Trends are descriptive, not causal ROI. Reconcile late evidence using versioned corrections and distinguish changed practice from changed measurement.
Act on critical policy effects immediately; do not wait for a reporting cycle. Use selection precision/recall at qualification, or service-level delivery and departmental outcome measures, only where relevant and their source links are feasible. Confirm purpose, lawful basis, access and retention before changing collection, following the [shared safeguards](./measurement/collection-and-safeguards.mdx).
</details>
**Next:** [Stage - 02 - Autonomous Agent →](./stage-02-autonomous-agent.mdx)
</div>