oss-sourcing
/oss-sourcing
Szuka darmowego API, zbioru danych lub biblioteki zamiast budowy od zera i sprawdza licencję, limity, bezpieczeństwo oraz utrzymanie.
- Autor
- devpilotX ↗
- Repozytorium
- ai-skill-ai ↗
- Licencja
- MIT
- Ostatnia zmiana repozytorium
- 31.10.2026
Ocena i bezpieczeństwo
- Pliki w paczce
- 5
- Skrypty
- 2
- Komentarze HTML
- 0
- Adresy URL w SKILL.md
- 1
Brak sygnałów. Repozytorium zmienione 31.10.2026, bez odwołań do wycofanych modeli i API.
MIT, otwarta: można używać i zmieniać, z zachowaniem informacji o autorze.
- 2 skrypty w paczce. Przejrzyj je przed instalacją, bo Claude może je uruchomić.
Skrypty: scripts/api_probe.py, scripts/repo_vet.py
Co robi i kiedy
Szuka darmowego API, zbioru danych lub biblioteki zamiast budowy od zera i sprawdza licencję, limity, bezpieczeństwo oraz utrzymanie.
Wyzwalacze z opisu
Oryginalny opis (EN)
Find a free API, dataset or library before building it, then vet it before depending on it. Use when the user needs weather, currency, geocoding, country or government data, asks "is there a free API for X" or "can we use this package", or is about to add a dependency or pick a data source. Checks licence and terms against the intended use, auth, rate limits, CORS, HTTPS, maintenance signals, security advisories, OpenSSF Scorecard, typosquatting and provenance, and sets a cache and fallback before any key leaves the server. Triggers on free API, public API, is there an API for, open dataset, w
Jak działa
- When to use and when to stay off
- Non-negotiables
- Procedure
- Self-audit
- What this cannot do
Podgląd SKILL.md
--- name: oss-sourcing description: Find a free API, dataset or library before building it, then vet it before depending on it. Use when the user needs weather, currency, geocoding, country or government data, asks "is there a free API for X" or "can we use this package", or is about to add a dependency or pick a data source. Checks licence and terms against the intended use, auth, rate limits, CORS, HTTPS, maintenance signals, security advisories, OpenSSF Scorecard, typosquatting and provenance, and sets a cache and fallback before any key leaves the server. Triggers on free API, public API, is there an API for, open dataset, which library, can we use this package, add a dependency, rate limit, licence check, supply chain, typosquatting. For one architecture or build-vs-adopt decision use arch-decide. For threat models and controls use security-hardening. For assembling UI from component kits use ui-compose. For runtime and framework versions and end-of-life use stack-select. For the full per-change verification loop use quality-gates. license: MIT metadata: version: 1.0.0 suite: ai-skill --- # Open-source sourcing Two failures sit at opposite ends of the same task. One is building a currency converter or a geocoder from scratch when a maintained free service already exists. The other is reaching for the first package that matches the name, pinning nothing, reading no licence, and finding out in production that the terms forbid the use, the project was archived two years ago, or the name was a typosquat. This skill sits between the search and the dependency: find what exists, then prove it is safe to depend on before anything imports it. ## When to use and when to stay off Run when the user needs external data or capability and a free or open option might exist (weather, exchange rates, geocoding, country reference data, government or statistical data), when they ask whether a specific API or package can be used, or when a dependency is about to be added and nobody has read its licence, limits or maintenance state. Stay off, and route instead, when: - The question is one architecture choice or build-versus-adopt with options to weigh. `arch-decide` owns the decision record; this skill feeds it the vetted candidates. - The work is a threat model or security controls for the whole system. `security-hardening` owns that; this skill covers only the sourcing of one dependency or service. - The need is UI components, icons, fonts or illustrations from kits. `ui-compose` owns that sourcing and its provenance ledger. - The question is which runtime or framework version to run and when it reaches end of life. `stack-select` owns version and support-window checks. - The work is wiring the per-change CI loop (format, lint, tests, secret and dependency scans). `quality-gates` owns that; this skill decides what to depend on, not how to gate every commit. The user says "stop", "I've decided", or "just execute": comply at once and stay off for the rest of the session unless asked again. ## Non-negotiables These override everything else in this file. 1. The licence and the terms of use are read against the intended use before the dependency is adopted. Commercial use, attribution, caching and redistribution are each checked, because a permissive code licence does not make a non-commercial data licence commercial. A terms page beats a repository licence file when the two disagree, and both are dated. 2. No figure is trusted from the owner's own page. Star counts, "millions of users" and uptime claims are read from the platform API or a status page this session, with the date, or labelled `ASSUMPTION:`. 3. Every link cited is opened this session. A link that does not resolve is removed with the claim it supported. 4. A dependency is pinned to an exact version or a commit, installed from a frozen lockfile, and its name is checked against the real package before install. A name one character or one hyphen away from a popular package is treated as a typosquat until shown otherwise. 5. Keys and tokens live server-side only, never in client code, a committed file or a log line. A service that forces the key into the browser is recorded as a client-exposure risk. 6. A rate limit, a terms change or an outage is planned for before launch: a cache layer, a pinned fallback source, and a documented switch path. A public donated service (Nominatim, Open-Meteo free tier) is never put directly behind user traffic without caching. 7. Where a licence carries a network or copyleft obligation (AGPL, GPL, the ODbL share-alike clause, a CC non-commercial term), the obligation is stated in plain words and whether the project can meet it is answered, not assumed. ## Procedure The source shortlist with licences and limits verified this session is in `references/source-registry.md`. The licence families and what each one demands are in `references/licence-matrix.md`. ### Step 1, decide whether to source at all Name what the code actually needs: the capability, the fields, the request volume at expected scale, and whether the use is commercial. A converter that needs daily exchange rates is a different problem from one that needs intraday rates, and the second rules out most free sources. If the need is trivial and stable (a static list of country codes), vendoring a small dataset once beats depending on a live service that can rate-limit or disappear. ### Step 2, find candidates from maintained lists Start from lists that are themselves maintained: `public-apis/public-apis` and APIs.guru for APIs, the relevant `awesome-*` list for a library ecosystem, and the official open-data portals for government and statistical data (data.gov.in for India, the World Bank and RBI DBIE for economic series). Treat every list entry as a lead, not a verdict: lists go stale, and an entry can point at a service whose terms changed. Carry two or three candidates into vetting, not one, so a failed vet does not end the search. ### Step 3, vet the licence and terms against the use Read the licence file and the terms-of-use page, and answer four questions in writing: may it be used commercially, does it require attribution and where, may responses be cached and for how long, and may the data be redistributed. Map the licence to its family with `references/licence-matrix.md`. A CC non-commercial term, an AGPL network clause or the ODbL share-alike obligation each changes what the product may do, so each is named with its consequence rather than waved through. ### Step 4, probe the service For an API, confirm the live facts rather than the documented ones: ``` python3 scripts/api_probe.py https://api.frankfurter.dev/v2/rates --tries 5 --json ``` It reports the status code, latency over N tries, whether HTTPS and a valid certificate are present, the CORS headers, and any rate-limit headers the service returns. Record the date of the run, because limits and terms change. For a service with a status page, read the recent uptime there rather than trusting an uptime claim on the marketing page. ### Step 5, vet the repository behind a library For a library or an open-source service, read the project's real health from the platform, not its README: ``` python3 scripts/repo_vet.py ossf/scorecard --json ``` It calls the GitHub REST API for stars, last push date, archived flag, licence SPDX id, open issue count and the latest release, and the OpenSSF Scorecard API for the aggregate security score when the project is covered. A token is read only from the `GITHUB_TOKEN` environment variable and never printed; without one the script still runs against the unauthenticated rate limit. Weigh the signals together: a high star count on a project last pushed three years ago and marked archived is not a safe dependency. ### Step 6, check the supply chain before install Confirm the package name character for character against the registry, not from memory, because a typosquat differs by a letter, a hyphen or a swapped scope. Prefer a package that publishes build provenance (npm provenance, a signed release) and pin it to an exact version or commit. Record the pinned version and the lockfile that installs it. Run the ecosystem's advisory scan (osv-scanner, `npm audit`, `pip-audit`) and read `quality-gates` for where that scan sits in CI. ### Step 7, record the decision and the exit Write a short source record: what was chosen, its licence and SPDX id, the limits and the date they were read, the attribution the product owes, the cache policy, the pinned fallback source, and the switch path if the service changes its terms or disappears. This record is what a future reader needs when the free tier moves behind a paywall, which it does without notice. ## Self-audit - The licence and the terms-of-use page were both read, dated, and answered against commercial use, attribution, caching and redistribution. - No star count, user count or uptime figure came from the source's own marketing page without a platform-API or status-page check. - Every cited link was opened this session; dead links and their claims were removed. - The chosen dependency is pinned to an exact version or commit and installs from a frozen lockfile. - The package name was checked character for character against the registry, and near-name collisions were ruled out. - No key or token appears in client code, a committed file or a log; server-side only. - A cache layer and a named fallback source exist for every live external service behind user traffic. - Any AGPL, GPL, ODbL share-alike or CC non-commercial obligation is stated with its consequence, not assumed away. - `api_probe.py` and `repo_vet.py` output carries the date it was run. ## What this cannot do It cannot read a terms-of-use page the site hides behind a login or a sales call; where terms are not public, it says so and treats the service as unverified. It cannot confirm a licence is compatible with the user's whole dependency graph. It checks one dependency at a time; a full licence-compatibility scan across the tree is a `quality-gates` concern with a dedicated tool. It cannot read private or paid registries, and `repo_vet.py` sees only what the GitHub API and the Scorecard API expose, not a project's private security posture. It does not give legal advice on licence obligations. Whether a specific use satisfies the AGPL network clause, the ODbL share-alike requirement, or a trademark restriction on an open-source logo is a legal question. Ask an intellectual-property or technology lawyer: "Given that we use this component under this licence, in this way, in this product distributed in these countries, what obligations do we take on and can we meet them as built?"
Pliki w paczce (5)
- SKILL.md 10 KB
- references/licence-matrix.md 5 KB
- references/source-registry.md 5 KB
- scripts/api_probe.py 11 KB
- scripts/repo_vet.py 13 KB
Wersje
W bazie nie ma innych skilli o nazwie oss-sourcing.
Migawka bazy z 13.11.2026 · wersja paczki 4b1c144f81cc (skrót SHA-256 całej paczki)
Pobrania w ostatnich 30 dniach: 0 · ranking
Zgłoś problem z tym skillem: szkodliwa treść, błędny opis albo prośba autora o usunięcie.
Podobne skille
- 86vet-dependency
Sprawdza pakiet przed dodaniem lub podbiciem: ostrzeżenia bezpieczeństwa, wiek wydania, popularność, licencję i dopasowanie. - 100open-data-sourcing
Szuka, sprawdza i cytuje otwarte dane GIS, np. działki, lidar, strefy zalewowe i granice, weryfikując licencję, zasięg i aktualizacje. - 100open-sourcing
Sprawdza gotowość projektu do publikacji open source: sekrety, licencję, dokumentację, CI i pakiety językowe. - 100gh-oss-release-prep
Prowadzi przez listę kontrolną przygotowania repozytorium do publikacji jako open source, aby projekt spełniał standardy OSS przed upublicznieniem. - 100free-apis-catalog
Podaje wyselekcjonowany katalog darmowych API ze zweryfikowanymi limitami planu darmowego, przy poszukiwaniu danych o pogodzie, wiadomościach, finansach lub sporcie. - 100assess-oss
Ocenia projekty open source na podstawie danych z GitHub, npm i crates.io: utrzymanie, popularność i licencje.