Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add instructions/alevizio/curb/claude-mdgit clone --depth 1 https://github.com/alevizio/curbWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/instructions/alevizio/curb/claude-md)<a href="https://agentmods.dev/instructions/alevizio/curb/claude-md"><img src="https://agentmods.dev/badge/instructions/alevizio/curb/claude-md.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.06277 | $0.06277 |
| Opus 5 | $0.03138 | $0.03138 |
| Sonnet 5 | $0.01255 | $0.01255 |
| Haiku 4.5 | $0.00628 | $0.00628 |
Grade A, and why
curb CLAUDE.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 6d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 315 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CURB — SF street parking, block by block
Context for Claude Code. Read this before editing.
What it is
A mobile-first web app that shows San Francisco street-parking rules on an interactive map: where you can park, until when it's swept, whether it's metered, and whether it's a residential permit (RPP) zone. Lets the user set a calendar reminder before the next sweep.
Stack (intentionally minimal)
- Single static file:
index.html. No build step, no framework, no bundler. - Vanilla JS + Leaflet 1.9.4 (from cdnjs) for the map.
- Basemap: official Google Map Tiles API when
GMAPS_KEY(orwindow.GMAPS_KEY) is set — session-token flow ininitBasemap(), viewport attribution refreshed on moveend. Falls back to keyless CARTO Voyager raster tiles when no key / on any failure. Leaflet stays the map engine either way. The Google key is a client key (referrer-restrict it) kept OUT of the public repo: local dev reads a gitignoredconfig.js(fromconfig.example.js); on Vercel,api/config.jsemitswindow.GMAPS_KEYfrom theGMAPS_KEYenv var andvercel.jsonrewrites/config.js→/api/config. - Everything is client-side. Data is fetched live from DataSF (Socrata) at runtime.
- Design system: fonts Anton (display) + Hanken Grotesk (body); "transit signage" aesthetic; color tokens in :root (--green clear / --amber soon / --red now / --meter permit-blue / paper+ink). Keep this language if extending the UI.
Data sources (all DataSF Socrata, CORS-open: access-control-allow-origin: *)
- Street sweeping —
yhqp-riqshttps://data.sfgov.org/resource/yhqp-riqs.json Fields: cnn (segment id), corridor, limits (cross streets), blockside, cnnrightleft (L/R vs digitized direction), weekday, fromhour, tohour, week1..week5 ("1"/"0" = Nth occurrence of that weekday in the month), line (GeoJSON LineString). CURRENT data. - Parking meters —
8vzz-qzz9https://data.sfgov.org/resource/8vzz-qzz9.json Fields: street_name (UPPERCASE), cap_color, on_offstreet_type, lat/long, etc. CURRENT data. Used only for a street-level count (no spatial join — see limits). - Parking regulations / RPP —
hi6h-neyhhttps://data.sfgov.org/resource/hi6h-neyh.json Fields: regulation, rpparea1 (permit-area letter), hrlimit, days, from_time, to_time, exceptions, shape (GeoJSON MultiLineString). STALE: this is SFMTA's 2017 set, flagged by the city as not comprehensively updated. Treat as a hint. NOTE: RPP covers BOTH curbs — rendered as ONE street-wide centerline ribbon UNDER the curb lines, with zoom-scaled weight (rppWeight(): 4px@z15 → 26px@z18 ≈ curb-to-curb). Do NOT draw offset bands per side: they stack into a blue blanket at low zoom and collide with the ±5m curb lines at high zoom (tried 2026-06-09, looked broken). - Loading / color-curb zones —
6cqg-dxku(Meter Operating Schedules) Fieldapplied_color_rulecarries the regulation + days_applied/from_time/to_time/ time_limit (White=passenger, Yellow=commercial, Red=truck, Green=short-term, Orange=bus).cap_coloris UNRELIABLE (white zones show Grey caps) — match on applied_color_rule. No geometry → join to meter coords bypost_id(8vzz-qzz9 lat/long). Metered zones only; paint-only curbs aren't published. Loaded once on toggle, rendered per-viewport. - Parking citations —
ab4h-6ztd(23.8M rows, daily, ~2-5 day lag) STR CLEAN (TRC7.2.22) + ST CLEANIN (T37C) = street-cleaning tickets, minute-resolution. A 2024 records request (#26-5453) restored GPS lat/long on the citations: ~815k of ~1M street-cleaning tickets now match to a CNN by NEAREST CNN SEGMENT (<=40m), the primary pipeline. The old address→CNN join via EAS (3mea-di5p, keyed by stripZeros(number)|street_name) is demoted to a pre-2024 fallback for rows without GPS. Precomputed offline intodata/enforcement.json— seescripts/build-enforcement-records.pyanddocs/sweeper-data-research.md. Powers the "🎯 Ticketed ~9:11a" lines.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 6d ago First seen · 315 lines · 6,277 tokens per session scan A 595130157dbe
curb CLAUDE.md is an instructions file published in the GitHub repository alevizio/curb (11 stars, last pushed 4d ago), licensed MIT. It adds 6,277 tokens to every session, about $0.0314 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other instructions, from other repositories
ph-civic-data-mcp CLAUDE.md
Claude Code instructions for xmpuspus/ph-civic-data-mcp, covering claude.md - ph-civic-data-mcp, what this is, contracts every change must hold, registration and module layout and psa openstat landmines.
indonesia-civic-stack CLAUDE.md
Instructions for suryast/indonesia-civic-stack, covering claude.md — instructions for claude code, mcp integration, quick context, before writing code and commands.
forthepeople CLAUDE.md
Claude Code instructions for jayanthmb14/forthepeople, covering claude.md — forthepeople.in project instructions, project, database schema changes (critical — changed june 2026), documentation location and completed prompts archive.
indonesia-civic-stack copilot-instructions.md
Instructions for suryast/indonesia-civic-stack, covering copilot instructions, key rules and file structure per module.
indonesia-civic-stack AGENTS.md
Instructions for suryast/indonesia-civic-stack, covering agents.md — ai agent development guide, project overview, architecture, key patterns and every module must follow this contract.
MobiAI-Core CLAUDE.md
Claude Code instructions for ArisGuimera/MobiAI-Core, covering mobiai — mobile development ai ecosystem, bootstrap and principles.