Doordash Budget

Commands doordash 26 Scanned by NVIDIA SkillSpector

View DoorDash spend vs caps, edit the dd-guard spending policy interactively, and reconcile the intent ledger against real order history

Install Command

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

View on GitHub

claude-code — doordash-budget

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.

2 observations

scanner score 27/100

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-budget.md:18
Session Persistence
Rogue Agent · doordash-budget.md:42

Learn more about NVIDIA SkillSpector.

Content

DoorDash Budget Manager

Manage the dd-guard spending policy: $ARGUMENTS

State: ~/.claude/dd-guard/limits.json (policy) and ~/.claude/dd-guard/ledger.jsonl (spend ledger). This command is the ONLY sanctioned way to modify limits.json — the doordash-spend-guard hook blocks direct command-line writes to it by design.

Current State

Policy: !cat ~/.claude/dd-guard/limits.json 2>/dev/null || echo "(no policy yet — dd-guard seeds defaults on first run)"
Headroom: !bash .claude/skills/doordash-spend-guard/scripts/dd-guard.sh status 2>/dev/null || echo "(dd-guard.sh not installed — install the doordash-spend-guard skill)"

Task

Parse $ARGUMENTS and run the matching subcommand:

`status` (default when no arguments)

Present the dd-guard status output as a friendly table: each cap, spend to date, headroom left. Note how many ledger entries are unreconciled intents and remind that subtotals exclude fees/tips.

`set`

Interactive policy editing:

Show the current policy values one by one.
Ask the user which value to change and to what (per_order_max, daily_max, weekly_max, monthly_max, cooldown_minutes, allowed_hours.start/end). Setting a cap to null removes it.
Confirm the exact new policy with the user before writing — read back the full JSON you intend to save.
Write ~/.claude/dd-guard/limits.json with the Write tool (the hook permits interactive edits through this command's flow, not shell redirection).

`reconcile`

Match unpaid intents against reality:

Run dd-cli order history and list recent orders.
For each ledger line with "status": "intent", check whether a matching order (by date/restaurant/amount proximity) appears in history:
- Match found → update the line to "status": "paid".
- Clearly abandoned (old intent, no matching order) → ask the user, then mark "status": "abandoned" so it stops counting against caps.
Rewrite ledger.jsonl preserving all other lines untouched (read-modify-write; keep it valid JSONL).
Report: N paid, M abandoned, K still pending.

Rules

Never change policy values without explicit user confirmation of the final JSON.
Never delete ledger history — reconciliation only flips status.
If the user asks to raise caps mid-session right after a blocked checkout, confirm they understand the block reason first (that friction is the point of the tool).