Borrowing it
Nothing to install: this file belongs to FreemapSlovakia/freemap-v3-react. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/FreemapSlovakia/freemap-v3-react/main/AGENTS.mdgit clone --depth 1 https://github.com/FreemapSlovakia/freemap-v3-reactWrote 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/freemapslovakia/freemap-v3-react/agents-md)<a href="https://agentmods.dev/instructions/freemapslovakia/freemap-v3-react/agents-md"><img src="https://agentmods.dev/badge/instructions/freemapslovakia/freemap-v3-react/agents-md/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/instructions/freemapslovakia/freemap-v3-react/agents-md"><img src="https://agentmods.dev/badge/instructions/freemapslovakia/freemap-v3-react/agents-md.svg" alt="Reviewed on agentmods" width="80" 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 | $0.06577 | $0.06577 |
| Opus 5.5 | $0.02631 | $0.02631 |
| Sonnet 5.5 | $0.01315 | $0.01315 |
| Haiku 4.5 | $0.00658 | $0.00658 |
Grade A, and why
freemap-v3-react AGENTS.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 yesterday.
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 — 142 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Project notes for AI agents
(This file is also exposed as CLAUDE.md via a symlink so Claude Code loads it.)
Reference docs
Deeper format/architecture references live in doc/. Consult them before designing changes in the same area:
- Architecture & code-layout reference — the feature-folder convention, the Redux store wiring (
rootReducer.ts/processors.ts), the processor middleware that owns all side effects, state persistence + rehydration, URL-hash sync, the global vs. per-feature i18n split, path aliases, and the layer/tool registries. Read this before adding a feature, a side effect, persisted state, or a new menu/tool/modal — it saves re-reading the tree. - Translation style — how each language addresses the user: formal everywhere it distinguishes (vykanie,
Sie,vous,Lei, önözés, vikanje), neutral Polish, neutral wording preferred, and the exceptions for button labels and search keywords. Read this before writing or translating any UI string. - Drawing feature export/import format mapping — GPX/GeoJSON write+read for drawing points/lines/polygons, the
fm:*lossless-shadow design, and the curated Garmin/OsmAnd icon dictionaries. Read this before changing anything insrc/features/export/or in the drawing-conversion processors. - Track data: import, elevation, colorize, export — the shared elevation-acquisition layer (
src/shared/elevation.ts:fetchElevations/enrichElevations/densifyAlong), the shared colorizers (src/shared/colorizers/,isAvailable-gated per feature), the render-only densified geometry that's never exported, the per-consumer elevation policy (route auto-fill vs. track prompt vs. tracking keep-recorded), and KML/KMZ/TCX import + lossless GeoJSON↔GPX transfer. Read this before touching elevation, colorize, the elevation chart, or track-file import/export across routePlanner/dataViewer/tracking/mapFeaturesExport. - What a history entry shows — how
src/app/url/entryContent.tsputs an entry's features back from link params, a snapshot, atrack-uid=/import-url=download or a saved map: reads tagged with the entry that asked and joined when already in flight, the entry's own content over the map's, remote tracks imported and then dropped from the URL, andentryLoading()holding back every writer. Read this before touching restore on reload or Back/Forward,mapsRestoreProcessor, ortrackStore.ts. - GPS recorder integration — how the PWA drives the standalone Android recorder APK over its
127.0.0.1HTTP API: the Android-only availability gate, the loopback/Local-Network-Access connection rules, the wire contract (protocol.tsis the single place to change it), the start-via-intent://flow, how the three failure causes are told apart, and the/track?since=+ SSE catch-up. Read this before touchingsrc/features/gpsRecorder/or anything that talks to the recorder. - Panorama viewer — the 360° terrain view: why the terrain service is addressed directly (it takes the bearer token, but bounds only what one request may cost — the client is what holds an account to its tier), why a render is only ever started by an explicit act and must abort the one in flight, the split between the store and
renderHolder, the off-thread decode and the canvas the picture is panned on, the label layout, and how a second label source plugs in via the distance buffer. Read this before touchingsrc/features/panorama/. - Viewshed overlay — the
vlayer over the terrain service'sPOST /viewshed: why it is a layer with a toolbar rather than a tool, the radius/scale pair the service validates together, what its alpha means, the render key that decides when Update appears, and the shared render queue. Read this before touchingsrc/features/viewshed/. - Weather radar layer — the animated
Roverlay over the two-feedcache.bigware.skAPI: why every request must set its ownreferrerPolicy(the upstream authenticates byReferer, and the app is servedno-referrer), the wire contract and per-feed zoom bands, the premium bands on the timeline and why locked frames stay on the track, the transient/persisted slice split, the polling processor's deliberate exceptions, and the per-frame tile-layer pool with its cross-fade and pan-drop rules. Read this before touchingsrc/features/weatherRadar/or the radar layer's tile source. - Tile attribution — what the outdoor map credits, decided by the tiles rather than by the countries in view: the dataset codes and the two ways they arrive (
Server-Timing: attrfor a tile on screen, since an<img>yields no response;X-Attributionfor anything holding one, such as the offline downloader), the two gates —Timing-Allow-OriginandAccess-Control-Expose-Headers— and how either one missing looks exactly like a tile reporting nothing, the painted-tile union and the three Leaflet traps in it,/licensesas the authority for what a code means, what is credited when no codes arrive, and the rules that keep an offline map's stored union honest. Read this before touchingsrc/shared/tileAttribution.ts,tileLicenses.ts,Attribution.tsx,exportCredits.ts, or the code capture incacheTilesProcessor.ts. - WebMCP agent tools — the tools
src/features/webMcp/registers ondocument.modelContextso a browser AI agent can drive the running app through the store: the API shape and its traps (it isdocument, notnavigator; secure context; thetoolspermissions policy), the lazy registration fromindex.tsx,defineTool's zod → JSON Schema wrapping, andwaitForStatefor the two tools that must await a processor's answer. Read this before adding or changing an agent tool. - SEO bot-prerender architecture — how crawlers get static HTML snapshots while humans get the SPA: the
sitemap-generator/(homepages, layer/tool hub pages, per-feature POI pages), the nginx/__prerenderrouting, and therspackindex-<lang>variants. Read this before changingsitemap-generator/, the sitemap/robots wiring, or the bot routing in the nginx vhost. - Self-hosted Photon geocoder — the search backend on fm5 that replaces
nominatim.openstreetmap.org: why the prebuiltphoton-db-*tarballs are unusable (four languages only) and we import the JSONL dump ourselves, the-languagesset and why widening it costs a re-import, the install layout, the systemd unit and nginx vhost (rate limit, 24 h cache,gzip_types), and the import/re-import runbook. Read this before touching the geocoder host, re-importing a dump, or pointing search at Photon. - Map library — every non-custom map is a library entry the user installs: the decisions (installed controls only what the UI offers, ids stable forever, library maps are not custom maps), the split registry (
src/shared/mapLibrary/mapIndex.tsx+ one body file per map indefs/, bundled vs lazily loaded bodies, themapLibraryslice and which selector to read), open questions, next steps (library modal, coverage union, harvesting) and the catalog research (ELI, JOSM, QMS, licences). Read this before touching layer definitions, how they load, install state, or the library. - Build & deploy gotchas — the
DEPLOYMENTenv var (not--mode) selects the prod build, native CSS nesting must be downleveled via LightningCSS, JS built-ins are core-js-polyfilled per use, and the nginx cache-header rules (entry HTMLno-store, only hashed assetsimmutable). Read this before changingrspack.config.ts, the CSS pipeline, or the nginx vhosts. - React Compiler — the app is compiled with the React Compiler (
jsc.transform.reactCompilerinrspack.config.ts): why ESLint exists here only for its diagnostics and parses with Babel rather than typescript-eslint, why the bail-out is per function (so manual memoization must stay wherever the compiler skips), the three ways it has actually broken this app (a render-time read of Leaflet/DOM/module state gets frozen, state that only forces a re-render stops working, a binding shadowing an intrinsic tag renames the tag), andscripts/react-compiler-check.mjs, which catches what reading the source cannot. Read this before removing auseMemo/useCallback, adding'use no memo', or debugging a component that renders stale values. - Events UX plan — where the events feature (branch
events, both repos) is heading and the steps to get there: an event is viewed in an event mode of its own, never loaded into the user's map; the read-only map rendering it rests on; participants, follows, discussion. Read this before working onsrc/features/events/, and tick its steps as they land. - UI conventions — what each react-bootstrap button
variantmeans by role (dark= dismiss/close,primary= main CTA,secondary= default neutral action,danger= destructive,outline-primary= toggle,warning= upsell). Read this before picking avariantfor a new<Button>. - User manual — the in-app manual's chapters (
src/manual/, English fallback,#show=manual/<chapter>), how to write them (UI labels verbatim, formal address, no emoji), the:FaIcon:syntax and the rule to show the UI's own glyph, the screenshot script (scripts/manual-screenshots.mjs: 2× capture, 8 px map margin, manifest sizes, the 820 px full screen), and moving a document into the manual. Read this before writing or changing a manual chapter or its screenshots. - Feature-announcement log — what has shipped and is worth posting about (with the commits that carry it), and what has already gone out on Facebook / Mastodon. Update the backlog when a user-visible feature lands, and move a topic to Posted once it goes out. Market-intro posts live in the roadmap's Outreach log instead.
- Product reality & market position — what Freemap actually is, written from the code rather than from marketing copy: exact coverage (46 countries; the 33 national high-res terrain datasets — per territory, so Germany counts as its 16 states — and the big markets without), the feature inventory, the business model (€15/yr, self-served ads shown to non-premium users, no user tracking), honest strengths and weaknesses (no mobile app, no community tour content), the competitor table with prices, and the positioning that follows. Read this before writing any promotional copy or choosing a market, and verify load-bearing claims against the code it cites.
- Italy outreach working file — the live state of the Italian push: venue research (which communities already know Freemap and must never be "introduced" again vs. which are cold), the finished Italian post drafts, forum-account status and blockers, and a "start here" handoff list. Read this before doing anything on Italy; log posted outcomes to the roadmap's Outreach log.
- Austria outreach working file — the German-language push: why Austria goes first (engagement and reusable copy — the terrain hook now covers all of Germany too), the gipfeltreffen.at warm channel and its three-year-old unanswered objection, the German post draft, and venue/registration state. Read this before working on AT/DE/CH.
- Foreign-growth promotion roadmap — data-driven plan (from Matomo country/language/referrer cuts) for growing usage and premium conversions outside Slovakia: which markets to target (Italy, Poland, Hungary, Austria, …), which UI languages to add next (Slovenian, then French), and the phased outreach steps. Read this before working on foreign promotion, deciding the next UI localization, or interpreting the audience analytics.
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.
- yesterday Changed · +1 lines · +89 tokens per session 416a92808bb9
- 6d ago Changed · +1 lines · +124 tokens per session 41a3588eb879
- 11d ago Changed · +3 lines · +249 tokens per session 72833f5e695f
- 21d ago Changed · +4 lines · +119 tokens per session 55d3201210c3
- 22d ago Changed · +4 lines · +145 tokens per session 157fed4798d2
- 1mo ago Changed · +1 lines · +150 tokens per session 90be143d88b1
- 1mo ago Changed · +8 lines · +255 tokens per session 4bbb1e98f591
- 1mo ago Changed · -10 tokens per session 7240a759870e
- 1mo ago Changed · +51 tokens per session b6954fc0e414
- 1mo ago First seen · 120 lines · 5,405 tokens per session scan A 6c3c91f0e014
freemap-v3-react AGENTS.md is an instructions file published in the GitHub repository FreemapSlovakia/freemap-v3-react (49 stars, last pushed yesterday), licensed Apache-2.0. It adds 6,577 tokens to every session, about $0.0263 per session on Opus 5.5. 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.
freemap-v3-react AGENTS.md compared with instructions from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| freemap-v3-react AGENTS.mdFreemapSlovakia/freemap-v3-react | Instructions file | 49 | yesterday | Apache-2.0 | A | 6,577 | Codex, OpenCode |
| vscode buildNext.instructions.mdmicrosoft/vscode | Instructions file | 195k | yesterday | MIT | A | 9,719 | GitHub Copilot |
| next.js AGENTS.mdvercel/next.js | Instructions file | 143k | today | MIT | A | 7,461 | Codex, OpenCode |
| langchain AGENTS.mdlangchain-ai/langchain | Instructions file | 149k | yesterday | MIT | A | 4,680 | Codex, OpenCode |
| vscode oss-third-party-notices.instructions.mdmicrosoft/vscode | Instructions file | 195k | yesterday | MIT | A | 5,516 | GitHub Copilot |
| vscode copilot-instructions.mdmicrosoft/vscode | Instructions file | 195k | yesterday | MIT | A | 3,093 | GitHub Copilot |
| deepseek-harness AGENTS.mddeepseek-ai/deepseek-harness | Instructions file | 274k | 2d ago | MIT | A | 4,313 | Codex, OpenCode |