agentleFS
pariveopariveo/salesforce-agent-skills/.claude/skills/release-review/SKILL.md

release-review

Skill2 starsChanged 31 days agoMITView on GitHub

Maintainers of this repository only. Reviews every plugin skill under skills/ against a new Salesforce release and the current official documentation, and writes a findings report before anything in a skill changes. Use when a Salesforce release or its preview release notes come out, when asked whether the skills are current, or to update them for a release.

7 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

5 rules. 0 are common, 3 are distinctive, and 2 are very specific to this project.

  • 0

    Common

    Most of the 6 edited versions we read keep these.

  • 3

    Distinctive

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Use a scratch org or a sandbox, never a production org, and name it on every sf command instead of relying on the CLI's default org.

    line 93
  2. Distinctive:

    Step 5. Sources: Write the inventory and work through it:

    line 96
  3. Very specific:

    Write the report to .release-review/<release>.md:

    line 138
  4. Distinctive:

    Don't edit the skills, their scripts or their examples during the review.

    line 166
  5. Distinctive:

    Step 8. Move the baseline: When every finding is either changed or recorded as "no change", update these together in one pull request:

    line 170

“Common” is measured: kept by at least 6 in 10 of the 6 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Release review

Each Salesforce release can add a Flow element, change a limit, move a help page or retire a feature. The skills state all of these, so a release a skill hasn't seen can make it recommend the wrong element or miss a better one. This review finds those gaps and reports them. It covers every folder under skills/ that holds a SKILL.md, in one pass against one baseline. The skills change afterwards, one finding per pull request, as CONTRIBUTING.md describes.

The review never states a platform fact from memory. Every finding cites a page on help.salesforce.com or developer.salesforce.com, or a deploy result.

Workflow

Copy this checklist into your working notes and tick it off:

- [ ] 1. Read the baseline in maintenance/release.json
- [ ] 2. Confirm the target release and its API version
- [ ] 3. Read the release notes of every release after the baseline
- [ ] 4. Diff the Flow schema (needs an org on the target release)
- [ ] 5. Re-check every source the skills cite
- [ ] 6. Validate the examples at the target API version (needs an org)
- [ ] 7. Classify the findings and write the report
- [ ] 8. After the changes land, move the baseline

Working files go in .release-review/, which git ignores. Name them after the target release, for example .release-review/winter-27.md.

Step 1. Baseline. maintenance/release.json names the release the skills were last reviewed against, its API version and the review date.

Step 2. Target. The target is the newest release whose release notes are published on help.salesforce.com, generally available or in preview. Take its API version from the release notes, or from "Api Version" in sf org display --target-org <alias> on an org that runs it. Write both sources into the report. If the target equals the baseline, there is no release to review, so re-check the sources only (step 5).

Step 3. Release notes. Open the release notes of every release after the baseline, up to the target. help.salesforce.com loads its content with JavaScript, so read it in a browser, because a plain HTTP fetch can come back empty. Read these sections:

  • Flow Builder, Flow Orchestration and the other Automation sections
  • New and changed metadata types, for Flow
  • Retirements and release updates that name Flow, Process Builder or workflow rules
  • Agentforce and Data 360 items only where they change a Flow element or action

Then read "Verify before asserting" in every SKILL.md under skills/. Where one lists a platform topic outside these sections, such as the order of execution, read the release notes on that topic too. Take the topics from each SKILL.md as it stands, because this review keeps no list of its own.

List every item with its title, its URL, what it changes and the skill it concerns.

Step 4. Schema. Download the Metadata WSDL from an org on the target release (Setup, API, Generate Metadata WSDL) into .release-review/, then:

node scripts/diff-flow-schema.mjs .release-review/metadata-<api>.wsdl --api <api>

It lists the Flow types, fields and enumeration values that were added, removed or retyped since the newest snapshot in maintenance/flow-schema/, and where the tag lists of the checker it names in its output differ from the schema. Find each addition in the Flow page of the Metadata API Developer Guide and in the release notes from step 3. With no snapshot yet, the script compares the checker's tag lists only.

Then open every script under skills/*/scripts/, found by that glob rather than from a list of names. Check each Flow type, field and enumeration value a script names against the diff. A script that uses a removed or retyped name is a Checker finding against its skill. The checker the script already compared needs no second pass.

Use a scratch org or a sandbox, never a production org, and name it on every sf command instead of relying on the CLI's default org.

Step 5. Sources. Write the inventory and work through it:

node scripts/source-inventory.mjs > .release-review/<release>-sources.md

Open each official source once and check every claim listed under it. A page that moved, a number that changed, a renamed feature and a page that is gone are each a finding. For each "Not verified" mark, search the documentation again, because a new page can settle it. For each source on another host, look for an official page that states the claim.

The maintainer script sources are the official pages CONTRIBUTING.md cites for the maintainer scripts. They back no skill claim. Check each against the command next to it in CONTRIBUTING.md. A mismatch is a finding against that script or CONTRIBUTING.md, not against a skill.

Step 6. Examples. With an org on the target release:

node scripts/validate-examples.mjs --target-org <alias> --api <api>

It deploys copies of every skill's examples raised to the target API version, as a check-only deployment. A failure is a finding, with the exact error.

Step 7. Findings and report. Record the skill each finding concerns, and sort it into one kind:

KindWhat it isWhat it leads to
CorrectionA claim no longer matches its sourceA reference edit and a CHANGELOG line
RetirementAn element, action or behaviour a skill recommends is deprecated or retiredA decision table or escape condition change, then the eval suite
New elementA newer element that replaces a legacy pattern a decision table doesn't coverA candidate table row with an example, a deploy and an eval case
CheckerA tag the checker doesn't know, or a minimum API version it should enforceA check-flow.mjs change
LinkA source moved or disappearedA link update
No changeChecked, and the skill is still rightA line in the report, so the next review sees it was checked

A decision table lists newer elements that replace a legacy pattern. A long-standing practice that the release didn't change is out of scope, and so is a feature that needs no Flow element, such as a Flow Builder UI change.

Write the report to .release-review/<release>.md:

# Release review: <target release> (API <n>.0)

Baseline: <release> (API <n>.0), reviewed <date>
Target confirmed from: <URL or command>
Org used: <alias and edition> | none (steps 4 and 6 not run)
Schema diff: <command and summary> | not run (<reason>)
Example validation: <command and result> | not run (<reason>)

## Skills
- <skill folder>: <n> findings | no change

## Findings

### <n>. <kind>: <one-line title>
Skill: <skill folder>
Source: <URL or deploy result>
Affects: <file:line, ...>
Change: <what to change in the skill>
Changes a recommendation: yes (run the eval suite) | no
Needs an example deploy: yes | no

## Checked without change
- <source or release note item>

Don't edit the skills, their scripts or their examples during the review. The report comes first, as in the skills' own audits. Once the maintainer agrees with the findings, each one that needs a change becomes its own pull request.

Step 8. Move the baseline. When every finding is either changed or recorded as "no change", update these together in one pull request:

  • maintenance/release.json: release, apiVersion and reviewedOn, one entry for the whole plugin
  • The "Currency" line under "Verify before asserting" in every SKILL.md under skills/
  • The API version a SKILL.md falls back to when there's no org, in every skill that states one
  • The review sentence under "Compatibility" in README.md
  • maintenance/flow-schema/<api>.json, saved with node scripts/diff-flow-schema.mjs .release-review/metadata-<api>.wsdl --api <api> --write
  • A CHANGELOG line under [Unreleased] that names the release reviewed

The smoke test fails while these disagree, and while the checker's tag lists differ from the newest schema snapshot.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.