LCORE-2118: Add Claude Code skill for Dependabot CVE triage

.claude/commands/file-dependabot-cves.md

@@ -0,0 +1,95 @@
+---
+name: file-dependabot-cves
+description: Fetch Dependabot alerts, cross-reference against LCORE Jira tickets, and file tickets for gaps
+---
+
+Audit Dependabot vulnerabilities for `$repo` (default: `lightspeed-core/lightspeed-stack`) and cross-reference them against existing LCORE Jira tickets.
+
+## Step 1: Fetch Dependabot alerts
+
+Fetch open Dependabot alerts using:
+
+```
+gh api "repos/$repo/dependabot/alerts" --paginate --jq '.[] | select(.state == "open") | {number, state, severity: .security_vulnerability.severity, package: .security_vulnerability.package.name, ecosystem: .security_vulnerability.package.ecosystem, summary: .security_advisory.summary, cve: (.security_advisory.cve_id // "N/A"), ghsa: .security_advisory.ghsa_id, created: .created_at, fixed_version: (.security_vulnerability.first_patched_version.identifier // "N/A")}'
+```
+
+If `$repo` is not provided, default to `lightspeed-core/lightspeed-stack`. Deduplicate results by (CVE + package name) when a CVE is present, or (GHSA ID + package name) when CVE is null/N/A. This prevents collapsing distinct GHSA-only advisories for the same package into a single entry.
+
+## Step 2: Present severity summary
+
+Present a summary table with counts by severity (Critical, High, Medium, Low), then a breakdown table grouped by package: Package | Severity | CVE | Summary | Fix Version.
+
+## Step 3: Search LCORE Jira for existing coverage
+
+!Requires JIRA Atlassian MCP
+
+Search LCORE Jira for existing tickets per affected package. Batch package names into OR clauses to minimize API calls:
+
+- By summary: `project = LCORE AND (summary ~ "pkg1" OR summary ~ "pkg2" ...)`
+- By CVE label: `project = LCORE AND labels in ("CVE-XXXX-XXXXX", ...)`
+
+Fields: `summary,status,assignee,priority,labels`. Limit: 50. Paginate if needed.
+
+## Step 4: Cross-reference and classify
+
+Cross-reference each Dependabot alert to its LCORE ticket(s) and classify as:
+- **Covered**: open/in-progress ticket exists
+- **Closed**: ticket done
+- **Missing**: no ticket
+
+Present:
+1. A coverage table: Vulnerability | Sev. | Dependabot # | LCORE Ticket(s) | Status | Assignee
+2. A **gaps table** listing only the missing vulnerabilities with their GitHub alert title (from `security_advisory.summary`), severity, CVE, and fix version
+3. Key findings: coverage ratio, unassigned high/critical items, duplicate tickets that could be consolidated
+
+## Step 5: Verify gaps
+
+For each gap, cross-reference in JIRA (full-text search by CVE ID) and GitHub (confirm alert is still open) to verify it is a real missing issue. Drop false positives (e.g., already-closed tickets, stale alerts).
+
+## Step 6: Ask user which gaps to file
+
+Ask the user:
+- Whether they want to create LCORE tickets for the missing vulnerabilities
+- Which severity levels to include (e.g. "only medium and above", "all", or specific ones)
+- The target fix version (look up available versions from `jira_get_project_versions` for LCORE)
+- The component to assign (look up available components from `jira_get_project_components` for LCORE)
+
+## Step 7: Fetch full advisory details and draft tickets
+
+For each vulnerability the user wants to file, fetch the full Dependabot advisory description:
+
+```
+gh api "repos/$repo/dependabot/alerts/$alert_number" --jq '{summary: .security_advisory.summary, description: .security_advisory.description, cve: (.security_advisory.cve_id // "N/A"), remediation: (.security_vulnerability.first_patched_version.identifier // "No fix available"), vulnerable_range: .security_vulnerability.vulnerable_version_range}'
+```
+
+Structure each ticket as:
+
+| Field | Value |
+|-------|-------|
+| **Project** | LCORE |
+| **Type** | Vulnerability |
+| **Title** | The original GitHub advisory summary (from `security_advisory.summary`) |
+| **Component** | As chosen by user |
+| **Fix Version** | As chosen by user |
+| **Labels** | The CVE ID (if available), Security |
+| **Description** | The full `security_advisory.description` from Dependabot, followed by `**Remediation:** Upgrade <package> to >= <fix_version>` (or "No upstream fix available yet" if no fix exists, including the vulnerable range). |
+
+Present all drafted tickets in a table to the user for review before creating them.
+
+## Step 8: Find the parent CVE epic
+
+Search for the parent epic: `project = LCORE AND issuetype = Epic AND summary ~ "CVE" AND summary ~ "lightspeed-stack" ORDER BY created DESC`. Pick the one whose fix version matches the user's chosen fix version. If ambiguous or none found, ask the user.
+
+## Step 9: Create tickets after user confirmation
+
+Only after the user explicitly confirms the drafts, create the tickets using `jira_create_issue` with:
+- `project_key`: LCORE
+- `issue_type`: Vulnerability
+- `summary`: the GitHub advisory title
+- `description`: as structured above
+- `components`: user's chosen component
+- `additional_fields`: `{"fixVersions": [{"id": "<version_id>"}], "labels": ["<CVE-ID>", "Security"], "parent": "<EPIC_KEY>"}`
+
+Omit `parent` only if the user chose to skip it.
+
+Report back the created ticket keys and links.