agentleFS
forcedotcomforcedotcom/salesforcedx-vscode/.claude/skills/dependabot-alerts/SKILL.md

dependabot-alerts

Skill1k starsChanged 27 days agoBSD-3-ClauseView on GitHub

Triage GitHub Dependabot security alerts into one-point GUS work items that bump the outermost consumer. User-invoked only.

1 other project uses a version of this.

Adopted by

Steady: 0 new projects took it in the last 4 weeks, none the 4 weeks before.thin data

Week of Sep 13: 0 new projectsWeek of Sep 20: 0 new projectsWeek of Sep 27: 0 new projectsWeek of Oct 4: 0 new projectsWeek of Oct 11: 0 new projectsWeek of Oct 18: 0 new projectsWeek of Oct 25: 0 new projectsWeek of Nov 1: 0 new projects
Sep 2026Oct 2026Nov 2026
New projects taking a version of this file each week, over the 8 weeks our latest run read: only 0 in all, so this is a hint, not a trend.

Most recently taken up by

Who's talking about forcedotcom/salesforcedx-vscode

The biggest 5 of 52 mentions linking to the repository that our crawl found. Each line is their own headline.

See all 52

Talked about on GitHub and Stack Overflow (52 times).

On GitHub · 50

45 more on GitHub

The file

We split this skill into the rules it gives an agent

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

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 5

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    User-invoked only: Never auto-fire.

    line 11
  2. Very specific:

    If the chosen bump crosses a major version (e.g.

    line 44
  3. Distinctive:

    Read the consumer's release notes / CHANGELOG for every major crossed (e.g.

    line 48
  4. Very specific:

    Only fixing version forces a breaking migration → the WI is not a simple bump.

    line 61
  5. Very specific:

    Before creating a WI for a bump, check for an existing Dependabot PR proposing the same bump (gh pr list --author 'app/dependabot' --state …

    line 65
  6. Very specific:

    When an alert is split across consumers: 1 of N for <GHSA> (each PR stands alone.

    line 90
  7. Very specific:

    Follow gus-cli/SKILL.md for create mechanics (fields, confirmation, temp-Subject + flags-dir flow) and runner identity.

    line 92
  8. Distinctive:

    Every open alert is one of: a created WI (with a lockfile-version verification line), skipped as unfixable, skipped as a green Dependabot …

    line 96

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “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

Dependabot Alerts

Triage open Dependabot security alerts into GUS work items. Output is work items only — no branches, bumps, or PRs. The auto-build-wi workflow claims each WI and does the build/PR.

User-invoked only. Never auto-fire.

Never

  • No npm overrides in package.json. Ever.
  • No blind lockfile bumps. Trace the dependency to its outermost consumer first.

Input

  • Bare invocation → list open alerts: gh api repos/forcedotcom/salesforcedx-vscode/dependabot/alerts --paginate --jq '[.[] | select(.state=="open")]' | jq -s add. --paginate is required — the API defaults to 30/page and silently returns only page 1 without it (e.g. 33 open alerts, not the ~14 page 1 shows).
  • User-specified package / GHSA / CVE → act on that alert only.

(Repo is a monorepo, workspaces packages/*. npm 11.)

Per alert: trace the thread

Find who pulls the vulnerable package in, working outward to a package.json.

  1. npm why <vuln-pkg> (or npm ls <vuln-pkg>) from repo root → every path to a direct dep declared in a package.json (root or a packages/*).
  2. For each path, the outermost consumer is the direct dep at the top of that path. One alert can have several.

Per consumer: pick the fix

Try in order; stop at the first that works:

  1. Bump the consumer in package.json. A newer version of the consumer resolves <vuln-pkg> to a patched version (consumer already shipped the fix). Edit the version in whichever package.json declares it. Best — preferred.
  2. pnpm update <consumer>. Consumer's existing semver range already allows a patched <vuln-pkg>, but the lockfile is stale. No package.json change. Second best.
  3. Unfixable — consumer's latest still pins the vulnerable version. Skip this path silently. No WI.

Pick the lowest consumer version that resolves <vuln-pkg> to a patched version, never latest — fewer majors crossed, fewer breaking changes to vet.

Vet major-version bumps

If the chosen bump crosses a major version (e.g. ^1.x → ^3.0.0), spawn one subagent per crossing bump to vet it before drafting the WI. One bump may span several majors (1→3 = two changelogs); independent bumps vet in parallel.

Each subagent's task:

  1. Read the consumer's release notes / CHANGELOG for every major crossed (e.g. 1→3 means read 2.0.0 and 3.0.0). There may be one or five.
  2. For each breaking change found, grep our code for the affected API/behavior and decide if it breaks us specifically.
  3. Return: does the bump break us? What breaks, where (file:line), and the lowest version that clears the vuln without a breaking change.

What the subagent looks for:

  • ESM-only ("type": "module"). Our .github/actions/* and most packages are CJS (tsc module: node16, require() output). An ESM-only dep can't be require()d from a CJS build — breaking migration, not a bump.
  • Dropped Node engine support — must satisfy the consuming context (.github/actions/* run using: node20; extensions target their own engine).
  • Transitive major bumps the consumer drags in (e.g. Octokit majors) — only breaking if our usage touches the changed API.

Act on the verdict:

  • Breaking change avoidable at a lower major that still clears the vuln → prefer it, cap the range below the breaking major (e.g. ^2.0.0, not ^3.0.0). Note the cap + reason in Details__c.
  • Only fixing version forces a breaking migration → the WI is not a simple bump. Put the subagent's findings (what breaks, file:line, required changes) in Details__c so auto-build-wi plans for it.

Dependabot dedup screen

Before creating a WI for a bump, check for an existing Dependabot PR proposing the same bump (gh pr list --author 'app/dependabot' --state open):

Dependabot PR stateAction
Green / mergeableSkip — let Dependabot merge it. No WI.
CI failingClose the Dependabot PR (gh pr close), then create the WI (manual AI loop is better than a broken Dependabot bump).
CI still runningWait — don't create a WI, don't close. Re-check later.
NoneProceed to create the WI.

Create the work items

One WI per consumer-bump, not per alert. If one alert needs three consumers bumped → three WIs → three PRs. Isolation: each PR's CI/regression runs against exactly one change, so a failure points at one bump.

Each WI:

  • 1 story point.
  • [ai-auto] at the front of Subject__c (e.g. [ai-auto] bump <consumer> to <ver> for <GHSA> (<vuln-pkg>)).
  • Assigned to the runner (you).
  • Epic: IDEx - Mandates and Updates a3QEE0000023Fm92AE. If that epic is closed (Health__c in Completed/Canceled), use the most recent open trust epic instead — query team epics, match by Name.

Details__c = the fix recipe so auto-build-wi can execute blind:

  • GHSA/CVE id, vulnerable package + affected range.
  • The fix: which package.json, which consumer, target version (case 1) or pnpm update <consumer> (case 2).
  • Verification: the exact version that should land in pnpm-lock.yaml after the build.
  • When an alert is split across consumers: 1 of N for <GHSA> (each PR stands alone; Dependabot closes the alert once all land).

Follow gus-cli/SKILL.md for create mechanics (fields, confirmation, temp-Subject + flags-dir flow) and runner identity. Show the draft and wait for confirmation before any sf data create record.

Completion criterion

Every open alert is one of: a created WI (with a lockfile-version verification line), skipped as unfixable, skipped as a green Dependabot PR, or waiting on a running Dependabot PR. Report the disposition of each.

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.