Doordash Lunch

Commands doordash 27 Scanned by NVIDIA SkillSpector

Team lunch orchestration via DoorDash CLI - build a group round, emit checkout with split table, track payer rotation

Install Command

npx claude-code-templates@latest --command doordash/doordash-lunch

View on GitHub

claude-code — doordash-lunch

Scanned by NVIDIA SkillSpector

Automated static analysis (no LLM) · v2.12.0 · 2026-10-17

Partial analysis: some checks could not run on this component.

1 observation

scanner score 21/100

SkillSpector looks for patterns such as shell commands, credential access, network calls or unpinned dependencies. An observation is something worth reading in the files before you install, not a confirmed vulnerability. Components that run code locally, like hooks, MCP servers and security tooling, naturally show more of them.

Agent Config Directory Access
Agent Snooping · doordash-lunch.md:18

Learn more about NVIDIA SkillSpector.

Content

Team Lunch

Group order operation: $ARGUMENTS

Works with the doordash-group-orders skill (install it alongside — it defines the full flows; this command is the quick entrypoint). State: team-food.json (roster), .dd/round-<date>.json (active round), .dd/rounds.jsonl (history).

Current State

Roster: !cat team-food.json 2>/dev/null || cat ~/.claude/dd-cli/team-food.json 2>/dev/null || echo "(no roster yet)"
Active round: !ls .dd/round-*.json 2>/dev/null | tail -1 || echo "(none)"
Rounds history: !wc -l < .dd/rounds.jsonl 2>/dev/null || echo "0" rounds recorded

Task

Parse $ARGUMENTS:

`order` (default)

Run Flow 1 of the doordash-group-orders skill: roster check → constraint-aware restaurant shortlist via dd-cli search → per-member choices (favorites first, one question max per member) → build cart recording person → cart-item-id in the round ledger → show the cart grouped by person.

If the user pastes a thread of requests, parse it (Flow 2) instead of interviewing members.

`checkout`

Show the grouped cart one final time (per-person subtotals).
Emit the checkout URL — via bash .claude/skills/doordash-spend-guard/scripts/dd-guard.sh checkout <cart-uuid> when doordash-spend-guard is installed, otherwise dd-cli order checkout-url --cart-uuid <cart-uuid>.
Print a share-ready split table (markdown, ready to paste into Slack): person, items, subtotal share.
After the human confirms payment: pull order_uuid from dd-cli order history, ask who paid, append {date, order_uuid, payer, split} to .dd/rounds.jsonl, and offer the fee-proration fallback ("tell me the final total and I'll prorate the difference").

`whose-turn`

Read .dd/rounds.jsonl, compute per-member (consumed − paid) balances, name the next payer, show the balance table so the answer explains itself.

`roster`

Show/edit team-food.json interactively: add/remove members, update hard constraints, allergens (ask severity explicitly), favorites, dislikes. Confirm the final JSON before writing.

Rules

Never guess an order for a member with allergens.
cart-item-ids only from real cart show output — attribution depends on it.
Splits are pre-fee until the user reports the final total; say so in the split table footer.