---
description: Turn a repeatable marketing, RevOps or analytics job into a tested skill, or vet a public skill
argument-hint: [job to package or skill to adapt or fix or skill folder or URL to vet]
---

Use the marketing-skill-authoring skill for this task: $ARGUMENTS

Read martech-context/ first if it exists, especially stack.md, metrics.md, naming-rules.md and decisions.md, then list the skills already installed for each client the team uses.
If $ARGUMENTS is empty or vague, ask whether this is a new skill, an adaptation, a fix, a vet or a library review. Then ask for 3 to 5 real past runs of the job, with names, emails and record or account IDs replaced by placeholders; for a fix, the prompts that failed to trigger or triggered wrongly; for a vet, the skill's source and commit.
Follow the skill's procedure in order and finish with its acceptance checks.
Draft outside every folder a client loads skills from, and test in a disposable project. Do not install, commit, publish or upload anything before the owner approves.
Test sessions get no write-capable MCP, app or connector tool, and batches of test sessions start only after the owner approves the run plan.
Keep every skill read-only by default: reads and local drafts only, with live changes written as a change list. Propose permission settings as a diff; never change them.
Never ask for or store credentials or customer data. Use fictional or redacted fixtures.
When vetting, never install the skill or run its scripts, and give the verdict the skill's thresholds require.
Hand back the scorecard, the skill folder and command file, the lint result, the eval record with trigger scores and permission evidence, and the registry row, or the vetting report.

If the marketing-skill-authoring skill is not installed, install it with: npx skills add https://www.martechdevs.com --skill marketing-skill-authoring -a claude-code