grep-build-spreadsheet

Build a Spreadsheet

Routes a tabular-data request to Grep's app-builder expert with output_type=spreadsheet — sugar for expert_id=app-builder + effort=build + "Build a spreadsheet of: " prefix. The result is an HTML spreadsheet (typically index.html with copyable cells + export-CSV button) in the job's workspace.

When to use this skill vs the others

User wants	Use
Spreadsheet / table / CSV / comparison matrix	/grep-build-spreadsheet (you are here)
Slidedeck / presentation	/grep-build-slidedeck
Interactive HTML app	/grep-build-app
Plain research report (with maybe an inline table)	/research

Concrete signals: "spreadsheet of", "table comparing", "CSV with", "compare X vs Y vs Z on dimensions A, B, C", "build me a financial model", "dataset of top N".

If the user just wants a one-off comparison table inside a written report (not interactive, no export), use /research instead — the build-mode tax (~$2, 10-15 min) isn't worth it for a single table you'd paste into markdown.

Auto-update check

node "$SCRIPTS_DIR/update-check.js" 2>/dev/null &

Prerequisites

Spreadsheet jobs run at effort=build (~$2.00 PAYG, 10-15 min).

node "$SCRIPTS_DIR/auth.js" status

If "authenticated": false, automatically invoke /grep-login.

Resolve script + resources paths

SCRIPTS_DIR="$(dirname "$(dirname "$(dirname "$(readlink -f "${CLAUDE_SKILL_DIR}/SKILL.md")")")")/scripts"
RESOURCES_DIR="$(dirname "$SCRIPTS_DIR")/resources"

Step 1: Tell the user up front

"Spreadsheet jobs take 10-15 minutes and cost about $2 (PAYG) or count against your subscription. I'll submit now and stream live updates."

Step 2: Clarify axes (if needed)

If the prompt names the axes ("compare Stripe, Adyen, Braintree on transaction fee, dispute fee, supported currencies"), skip clarification.

If axes aren't obvious, use AskUserQuestion with two questions:

• Rows — what are the entities being compared? (vendors / companies / time periods / regions / products / etc.)
• Columns — what dimensions? (price / features / latency / SLA / market share / etc.)

Skip if the prompt is already specific.

Step 3: Decide whether to use a structured schema

If the user wants enforced column types (currency, percent, URL, etc.) or a citation policy per row, plan to pass --json-schema-file="$RESOURCES_DIR/spreadsheet_schema.json" to the submit command in Step 5 (do NOT submit yet — refine the prompt in Step 4 first).

The schema constrains output to {title, description, columns: [{key, label, type, unit}], rows: [{values, citations}], footer_notes}. Column types: string | number | currency | percent | date | url | boolean. Note: per-value type validation against the column types is left to the app-builder — the schema only enforces the shape (columns array, rows array, presence of required top-level fields).

For free-form spreadsheets (no enforced types), omit the schema flag.

Don't submit here. This step decides whether the schema flag will be added to Step 5's command. The actual submit happens once, in Step 5.

Step 4: Refine the prompt

Refine the user's raw request into a spreadsheet brief:

• Subject (what the spreadsheet is about)
• Rows (specific entities, ideally count-bounded — "top 10", "5 named providers")
• Columns (specific dimensions with units — "transaction_fee_percent", "p99_latency_ms")
• Sort / filter / interaction (if any — sortable, "best value" highlight, region filter)
• Citation policy (every row's value cited, or just key cells)

Example:

• Raw: "compare LLM providers"
• Refined: "Spreadsheet comparing the top 5 LLM API providers (Anthropic, OpenAI, Google, Mistral, xAI) on: input cost ($/M tokens), output cost ($/M tokens), max context (tokens), p99 latency (ms), tool-use support (boolean), vision support (boolean). One row per provider, sortable columns, source URL on every cell. Include a footer note listing the date of the pricing snapshot."

Step 5: Submit — use Monitor (background)

Submit ONCE here, using the refined prompt from Step 4. Append --json-schema-file=... only if Step 3 said to use the schema.

Free-form (no schema):

node "$SCRIPTS_DIR/grep-api.js" run "<refined>" \
  --output-type=spreadsheet --max-wait=1800 2>&1

Schema-constrained:

node "$SCRIPTS_DIR/grep-api.js" run "<refined>" \
  --output-type=spreadsheet \
  --json-schema-file="$RESOURCES_DIR/spreadsheet_schema.json" \
  --max-wait=1800 2>&1

--output-type=spreadsheet is sugar for --expert-id=app-builder --effort=build. Don't pass both.

Run with Monitor (timeout_ms: 1800000, persistent: false). Only one submit per skill invocation — the build tier is ~$2 each.

Step 6: Tell the user

"Spreadsheet job <slug> started — 10-15 min. I'll stream updates and post the URL when ready."

Step 7: List workspace + present URL

When the job completes:

node "$SCRIPTS_DIR/grep-api.js" files <slug>

Find the spreadsheet — typically index.html (sometimes with companion data.csv or data.json). Print the URL using $GREP_API_BASE (or fall back to https://api.grep.ai) so staging / preview environments work too:

API_BASE="${GREP_API_BASE:-https://api.grep.ai}"
echo "$API_BASE/api/v2/research/<slug>/files/index.html"

Tell the user how to use it:

"Your spreadsheet is ready: $API_BASE/api/v2/research/<slug>/files/index.html

• Sort any column by clicking the header
• Export CSV — most templates include a 'Download CSV' button. If not, copy cells directly into Excel/Sheets.
• Local copy: curl -L \"$API_BASE/api/v2/research/<slug>/files/index.html\" -H 'Authorization: Bearer <token>' > /tmp/spreadsheet.html && open /tmp/spreadsheet.html"

If a data.csv or data.json is in the workspace, mention it — those are easier to pipe into other tools than scraping the HTML.

Step 8: Follow-ups

If the user wants more rows, different columns, or a re-rank, use /grep-continue <slug> "<follow-up>" instead of a fresh job. The existing workspace + research context get reused — cheaper.

Anti-patterns

• Do NOT use this skill for static reports with embedded tables. Use /research and let it write a markdown table inline. Build mode is only worth it for interactive spreadsheets the user will sort / filter / export.
• Do NOT pass --effort=low or --effort=medium with --output-type=spreadsheet — sugar already pins effort=build.
• Do NOT pass --expert-id=app-builder alongside --output-type=spreadsheet — redundant; sugar already does that.
• Do NOT block on --max-wait=540 — spreadsheet builds need 10-15 min. --max-wait=1800 + Monitor.
• Do NOT submit without telling the user the cost + duration.
• Do NOT skip column-type enforcement when types matter (currency, percent, dates) — load the JSON schema. Free-form output may pick string types for everything, which breaks downstream sort/filter.
• Do NOT trust the URL is publicly accessible — workspace files require auth.

If the job times out

Exit code 2 means --max-wait elapsed but the job is still running server-side. Tell the user "Still building (job: ). Checking back." Use result <slug> once enough time passes, or invoke the /ultra-research polling pattern.