Claude GitHub Copilot Skill

assessing-change-impact

Assessing what a ServiceNow change request touched and what followed it: affected CIs, other changes on the same CIs in an overlapping window, and incidents on those CIs after the change, without trusting hidden incident-to-change links. Use this skill whenever the user asks whet

LLM Mart · 0 points · 2 views 139 listing impressions 0 install-command copies
Download jschuller-mcp-server-servicenow-examples_change-impact_skills_assessing-change-impact-09e5cbc.zip · 2 KB

Install

skills CLI npx skills add https://github.com/jschuller/mcp-server-servicenow/tree/main/examples/change-impact/skills/assessing-change-impact
Claude Code claude plugin marketplace add https://llmmart.ai/marketplace.json && claude plugin install jschuller-mcp-server-servicenow@llmmart
Git git clone https://github.com/jschuller/mcp-server-servicenow.git

The skills CLI installs just this skill, for any of its supported agents. Claude Code installs the whole jschuller/mcp-server-servicenow collection as a plugin from our marketplace. Git is the plain clone.

Skill manifest

Assessing Change Impact

Read-only evidence for a post-implementation review: what a change touched, what overlapped with it, and what broke afterwards. It finds correlation, not cause. Say so in the report, and let the people who ran the change judge.

Tables and roles

Table What it gives Readable with
change_request The change: CI, planned window (start_date, end_date), state, type, risk itil
task_ci Affected CIs of a task (task, ci_item) itil
incident Incidents by CI and time itil

When access is denied

  • A 403 on any of the three tables: the review is not assessable with this identity; say which table.
  • Hidden incident fields. The direct links incident.rfc ("Change Request") and incident.caused_by ("Caused by Change") are often hidden by ACLs. A query on a hidden field is ignored, not rejected: it returns every incident. Check visibility first (workflow step 4), and fall back to the CI-and-time correlation.
  • 0 rows means nothing was found with this identity. Rows may be hidden, so don't call a change clean on that alone.

Workflow

Progress checklist:

- [ ] Read the change and its window
- [ ] List its affected CIs
- [ ] Find other changes on those CIs in an overlapping window
- [ ] Check whether incident→change links are readable
- [ ] Find incidents on those CIs after the change
- [ ] Report evidence, separating facts from judgement
  1. Read the change:

    list_records(table_name="change_request", query="number=CHG0000076", fields="sys_id,number,short_description,cmdb_ci,start_date,end_date,state,type,risk", limit=1)
    
  2. List its affected CIs. The main cmdb_ci is usually one of them, but not always:

    list_records(table_name="task_ci", query="task.number=CHG0000076", fields="ci_item,ci_item.name", limit=100)
    
  3. Overlapping changes on the same CI. A change whose planned window starts before this one ends, and ends after this one starts, overlaps it. Put the change's own window into the query as gs.dateGenerate literals:

    list_records(table_name="change_request", query="cmdb_ci=<ci_sys_id>^start_date<javascript:gs.dateGenerate('2026-08-07','15:00:00')^end_date>javascript:gs.dateGenerate('2026-08-07','14:30:00')^number!=CHG0000076", fields="number,short_description,start_date,end_date,state", limit=50)
    

    Changes that list the CI only as an affected CI show up through task_ci instead:

    list_records(table_name="task_ci", query="ci_item=<ci_sys_id>^task.sys_class_name=change_request", fields="task.number", limit=100)
    
  4. Can this identity read the direct links? Request them on a few incidents:

    list_records(table_name="incident", fields="number,rfc,caused_by", limit=5)
    

    If the result carries a warning naming rfc or caused_by, the links are hidden: skip step 5a.

  5. Incidents after the change:

    • a. Where the links are readable: list_records(table_name="incident", query="caused_by=<change_sys_id>^ORrfc=<change_sys_id>", fields="number,short_description,priority,opened_at", limit=50).
    • b. In every case, incidents on each affected CI opened in the 48 hours after the planned end:
      list_records(table_name="incident", query="cmdb_ci=<ci_sys_id>^opened_atBETWEENjavascript:gs.dateGenerate('2026-08-07','15:00:00')@javascript:gs.dateGenerate('2026-08-09','15:00:00')", fields="number,short_description,priority,opened_at,state", limit=50)
      
  6. Report:

    • the change and its window;
    • its affected CIs;
    • overlapping changes, a collision risk worth raising before the next window;
    • incidents found by direct link, and separately the incidents found by CI and time.

    Label the second group as correlation. Say which checks were not assessable.

Tips

  • Dates in queries are in the caller's time zone, which matches the values the Table API returns for that identity.
  • Widen the 48-hour window for changes that take effect slowly (caches, batch jobs), and say you did.
  • ServiceNow's own conflict detection compares planned windows and CIs too, and also blackout and maintenance schedules. Where it runs, its results are on the change; this skill is the read-only, cross-change view.
  • Leave conclusions about cause to the change owner; offer the evidence.

Sources

Files (mcp-server-servicenow)
  • SKILL.md 6.4 KB
    ---
    name: assessing-change-impact
    description: "Assessing what a ServiceNow change request touched and what followed it: affected CIs, other changes on the same CIs in an overlapping window, and incidents on those CIs after the change, without trusting hidden incident-to-change links. Use this skill whenever the user asks whether a change (a CHG number) caused or broke anything, about a post-implementation review, change collisions or conflicts, or what else is scheduled on a CI, even if they do not ask for a skill."
    license: MIT
    metadata:
      author: mcp-server-servicenow examples
      version: "0.1.0"
      io.github.jschuller/servicenow-roles: "itil snc_read_only"
      io.github.jschuller/servicenow-tables: "change_request task_ci incident"
    ---
    
    # Assessing Change Impact
    
    Read-only evidence for a post-implementation review: what a change touched, what overlapped with it, and what broke afterwards. It finds correlation, not cause. Say so in the report, and let the people who ran the change judge.
    
    ## Tables and roles
    
    | Table | What it gives | Readable with |
    |-------|---------------|---------------|
    | `change_request` | The change: CI, planned window (`start_date`, `end_date`), state, type, risk | `itil` |
    | `task_ci` | Affected CIs of a task (`task`, `ci_item`) | `itil` |
    | `incident` | Incidents by CI and time | `itil` |
    
    ## When access is denied
    
    - A 403 on any of the three tables: the review is not assessable with this identity; say which table.
    - **Hidden incident fields.** The direct links `incident.rfc` ("Change Request") and `incident.caused_by` ("Caused by Change") are often hidden by ACLs. A query on a hidden field is ignored, not rejected: it returns every incident. Check visibility first (workflow step 4), and fall back to the CI-and-time correlation.
    - 0 rows means nothing was found with this identity. Rows may be hidden, so don't call a change clean on that alone.
    
    ## Workflow
    
    **Progress checklist:**
    ```
    - [ ] Read the change and its window
    - [ ] List its affected CIs
    - [ ] Find other changes on those CIs in an overlapping window
    - [ ] Check whether incident→change links are readable
    - [ ] Find incidents on those CIs after the change
    - [ ] Report evidence, separating facts from judgement
    ```
    
    1. Read the change:
       ```
       list_records(table_name="change_request", query="number=CHG0000076", fields="sys_id,number,short_description,cmdb_ci,start_date,end_date,state,type,risk", limit=1)
       ```
    2. List its affected CIs. The main `cmdb_ci` is usually one of them, but not always:
       ```
       list_records(table_name="task_ci", query="task.number=CHG0000076", fields="ci_item,ci_item.name", limit=100)
       ```
    3. **Overlapping changes on the same CI.** A change whose planned window starts before this one ends, and ends after this one starts, overlaps it. Put the change's own window into the query as `gs.dateGenerate` literals:
       ```
       list_records(table_name="change_request", query="cmdb_ci=<ci_sys_id>^start_date<javascript:gs.dateGenerate('2026-08-07','15:00:00')^end_date>javascript:gs.dateGenerate('2026-08-07','14:30:00')^number!=CHG0000076", fields="number,short_description,start_date,end_date,state", limit=50)
       ```
       Changes that list the CI only as an affected CI show up through `task_ci` instead:
       ```
       list_records(table_name="task_ci", query="ci_item=<ci_sys_id>^task.sys_class_name=change_request", fields="task.number", limit=100)
       ```
    4. **Can this identity read the direct links?** Request them on a few incidents:
       ```
       list_records(table_name="incident", fields="number,rfc,caused_by", limit=5)
       ```
       If the result carries a `warning` naming `rfc` or `caused_by`, the links are hidden: skip step 5a.
    5. Incidents after the change:
       - **a.** Where the links are readable: `list_records(table_name="incident", query="caused_by=<change_sys_id>^ORrfc=<change_sys_id>", fields="number,short_description,priority,opened_at", limit=50)`.
       - **b.** In every case, incidents on each affected CI opened in the 48 hours after the planned end:
         ```
         list_records(table_name="incident", query="cmdb_ci=<ci_sys_id>^opened_atBETWEENjavascript:gs.dateGenerate('2026-08-07','15:00:00')@javascript:gs.dateGenerate('2026-08-09','15:00:00')", fields="number,short_description,priority,opened_at,state", limit=50)
         ```
    6. Report:
       - the change and its window;
       - its affected CIs;
       - overlapping changes, a collision risk worth raising before the next window;
       - incidents found by direct link, and separately the incidents found by CI and time.
    
       Label the second group as correlation. Say which checks were not assessable.
    
    ## Tips
    
    - Dates in queries are in the caller's time zone, which matches the values the Table API returns for that identity.
    - Widen the 48-hour window for changes that take effect slowly (caches, batch jobs), and say you did.
    - ServiceNow's own conflict detection compares planned windows and CIs too, and also blackout and maintenance schedules. Where it runs, its results are on the change; this skill is the read-only, cross-change view.
    - Leave conclusions about cause to the change owner; offer the evidence.
    
    ## Sources
    - [MobileGlideRecord - Scoped](https://www.servicenow.com/docs/r/brazil/api-reference/MobileGlideRecordScopedAPI.html) · [source](https://github.com/ServiceNow/ServiceNowDocs/blob/753a7f0e8fac/markdown/api-reference/MobileGlideRecordScopedAPI.md)
    - [GlideSystem - Global](https://www.servicenow.com/docs/r/brazil/api-reference/server-api-reference/c_GlideSystemAPI.html) · [source](https://github.com/ServiceNow/ServiceNowDocs/blob/753a7f0e8fac/markdown/api-reference/server-api-reference/c_GlideSystemAPI.md)
    - [Associated CIs on a change request](https://www.servicenow.com/docs/r/brazil/it-service-management/change-management/c_AffectedCIsAndImpactedServices.html) · [source](https://github.com/ServiceNow/ServiceNowDocs/blob/753a7f0e8fac/markdown/it-service-management/change-management/c_AffectedCIsAndImpactedServices.md)
    - [Conflict detection](https://www.servicenow.com/docs/r/brazil/it-service-management/change-management/c_ConflictDetection.html) · [source](https://github.com/ServiceNow/ServiceNowDocs/blob/753a7f0e8fac/markdown/it-service-management/change-management/c_ConflictDetection.md)
    - [Create an incident](https://www.servicenow.com/docs/r/brazil/it-service-management/incident-management/create-an-incident.html) · [source](https://github.com/ServiceNow/ServiceNowDocs/blob/753a7f0e8fac/markdown/it-service-management/incident-management/create-an-incident.md)
    

Comments (0)

Sign in to join the conversation.

No comments yet.

Reviews (0)

No reviews yet.