Edit Google Apps Script locally with Claude Code and clasp push
Clone an Apps Script project with clasp, open Claude Code in that folder, describe the change in plain English, review the diff against your existing patterns, then clasp push — the same loop jlhernando uses for SEO reporting scripts.
Tools used
- Claude CodeReads the local Apps Script project and drafts changes that match existing patterns
- claspClones the Apps Script project locally and pushes reviewed edits back
- Google Apps ScriptRuntime for the Sheets/BigQuery automation Hernando maintains
Step navigator
A local clasp checkout you can edit with Claude Code, a reviewed change that matches your existing helpers/menus, and a successful clasp push back to the Apps Script project.
SEOs and automation folks who already have Apps Script projects and want a local Claude Code workflow instead of the web editor.
Before you start
José Luis Hernando (jlhernando, 15 Jan 2026) maintains ~1,500-line Apps Script SEO reporters that query BigQuery Bulk Export, build weekly trend sheets, and email exec summaries. The browser editor was too painful to stay in; clasp + Claude Code became his loop: clone once, describe the gap, review, push. This guide follows his post only. His exact Script IDs, SQL builders, and full catchUpMissingWeeks source stay unpublished gaps.
Step-by-step instructions
Clone the Apps Script project once with clasp
Hernando's new workflow step 1: clone locally with your IDE (he mentions VS Code) using clasp clone <scriptId> — only needed once per project. Find the Script ID in the Apps Script editor under Project Settings.
Why this works
Without a local checkout Claude Code cannot read your real helpers and menus.
Open Claude Code in the project folder and describe the gap in plain English
Step 2–3 in his loop: open Claude Code in the project directory with claude, then describe what you want. Real example after holidays: Week 1 of 2026 was missing from the weekly reporter, so he asked for a function that checks for missing weeks and starts from the gap. Within a minute Claude Code had read Code.js, understood ISO weeks / BigQuery / trend sheets, written catchUpMissingWeeks(), and added it to the menu — reusing his existing backfillTrendWeeks() pattern and toast style.
I'm starting this in week 2 of Jan, can I have a function that checks if weeks are missing and starts from the week missing?
Why this works
Plain-English gaps beat hand-writing BigQuery date SQL in the web editor.
Use it to debug mismatches you notice after a run
After catch-up he saw data out to column AK while headers only reached Y. He told Claude Code; it compared getTrendHeader() (25 columns) to newer SQL output (37 columns), found missing Brand / Non-brand Impressions headers, and fixed the header array — work he estimates at ~20 minutes of manual counting.
I'm getting data until AK column but headers until Y column - check what's the problem
Why this works
Describe the symptom; let it cross-check your header helper against the SQL aliases.
Review the diff, then clasp push — tips he actually uses
Final step: review and clasp push. His getting-started tips: be specific about the stack (or /init); let it read before it writes; paste errors for debugging; trust but verify, especially before anything that touches production data. For complex features he asks Claude Code to plan first; for quick fixes he just describes the problem.
Why this works
Push without review is how production Sheets get a confident wrong header row.
End of guide · Reviewed Nov 1, 2026
Pro tips2
- Tell Claude the stack up front (Apps Script + BigQuery) or run /init so it stops guessing APIs.
- Let it read the file before it writes — Hernando says interrupting that read costs context.
Common mistakes2
- clasp push without reviewing — he still treats production data scripts as trust-but-verify.
- Asking for a feature without mentioning existing helpers; Claude Code matched his backfillTrendWeeks pattern only because it read the repo.
FAQ2
Is this only for SEO scripts?
Do I need the Apps Script web editor afterward?
Related field guides.
Stop Claude Code repeating the same mistakes with mechanical PreToolUse guards
Install the zero-config self-kill Bash guard as a Claude Code plugin (or copy the seven hooks/tools), merge the hooks block into settings.json, turn optional rules on only with real values, then prove a blocked command with their test suite — each guard exists because an agent made that mistake more than once.
Check and remediate the backdoored Claude Code extension from the Cursor marketplace
Confirm whether you installed Anthropic.claude-code 2.1.312 or 2.1.313 from Cursor's marketplace on 29 Oct, grep for the svc-telemetry SSH key, run Cursor's remediate scripts if hit, then update to 2.1.314 and rotate credentials from a clean machine.
Teach Claude Code a security-response skill with live-write guards
Write a Claude Code skill that inventories your estate from GitHub, fans out exposure/mitigation/patch/rollout agents, labels every claim verified/read/assumed, proves fixes in capped containers, and blocks live writes behind a PreToolUse ask hook.
Measure Claude Code agent_summary spend on background-agent status labels
Upgrade past the cache-miss fix, turn on OpenTelemetry, group api_request events by query_source, then shrink or re-route Agent-tool subagents so status-label calls stop eating your quota.
Automate unstructured repo work with GitHub Agentic Workflows
Install gh aw, initialise the repo, have a coding agent write a Markdown workflow, compile it to lock.yml, commit both, and put a drift check in CI so you never ship Markdown without its compiled Actions file.
Reduce attack surface with Claude before you chase individual CVEs
Have Claude inventory what code is reachable by default, before authentication, and under conflicting config options, then gate or feature-flag what you do not need, and prove the change with the project tests.
Grow a shared CLAUDE.md from every PR mistake
Check one CLAUDE.md into git, add a rule whenever Claude repeats a mistake, tag @claude on review to capture learnings, start in Plan mode, and give Claude a way to verify its own work.
Ship AI-written code you can still explain: own the design, make the agent show its assumptions
Decide the abstractions yourself, pin them in a sketch or diagram, have the agent describe every change and flag what it assumed, then ask it to explain the diff before you commit.