Przejdź do treści
Biblioteka skilli
Programowanie · Analiza

dependency-add

/dependency-add

Pomaga dobrać nową zewnętrzną zależność do projektu lub wybrać między kandydatami bibliotek, pakietów i narzędzi.

Repozytorium
skills ↗
Licencja
MIT
Ostatnia zmiana repozytorium
03.11.2026

Ocena i bezpieczeństwo

100/ 100Klasa: dobry
Format25 / 25
Frontmatter z nazwą i opisem, poprawna nazwa, opis do 1024 znaków, nazwa zgodna z folderem.
Odkrywalność25 / 25
Opis mówi, kiedy użyć skilla, ma sensowną długość i jest w 3. osobie.
Treść35 / 35
Co najmniej 150 słów i do 500 linii, nagłówki, listy kroków, przykłady albo kod.
Higiena15 / 15
Bez zaślepek typu TODO, bez ścieżek Windows, z licencją.
Ryzykobrak sygnałów w skanie
Pliki w paczce
2
Skrypty
0
Komentarze HTML
0
Adresy URL w SKILL.md
0
Przestarzałość

Brak sygnałów. Repozytorium zmienione 03.11.2026, bez odwołań do wycofanych modeli i API.

Licencja

MIT, otwarta: można używać i zmieniać, z zachowaniem informacji o autorze.

Co robi i kiedy

Pomaga dobrać nową zewnętrzną zależność do projektu lub wybrać między kandydatami bibliotek, pakietów i narzędzi.

Oryginalny opis (EN)

Use whenever adding a new third-party dependency to any project, or when choosing between candidate libraries, packages, or tools to depend on.

Jak działa

  1. Objective
  2. Process
  3. Fast Checks With gh
  4. Approval
  5. Stop On Malware
  6. Output
Sekcje z nagłówków SKILL.md, w oryginale · 77 linii, 723 słów
Podgląd SKILL.md
---
name: dependency-add
description: Use whenever adding a new third-party dependency to any project, or when choosing between candidate libraries, packages, or tools to depend on.
---

# Dependency Add

## Objective

Every new third-party dependency is vetted before it is added or installed. A candidate that meets the trust bar is added without asking. Anything else waits for the user's explicit approval, and a candidate with known malware stops all work.

Dependencies we own, such as the user's own packages, libraries, and flake inputs, are out of scope. Add those normally.

Do not add a dependency merely for convenience.

## Process

1. **Find candidates.** List the best candidates from knowledge, or find them with `/research-web`.
2. **Filter with `/dependency-judge`.** Drop candidates that lose to keeping the code local, or that fail its maintenance, license, and transitive-risk checks.
3. **Check security.** Start with the advisory queries in Fast Checks With gh, then use `/research-web`. For each remaining candidate, at the exact version you plan to use, search for actively exploited vulnerabilities, major CVEs, supply-chain compromises, and malware or infostealers reported or trending in the last full month. Check the package itself and any compromise of its registry account or maintainers.
4. **Apply the trust bar.** A candidate meets it when its upstream repository has more than 1,000 GitHub stars, verified with `gh repo view`, and step 3 found no major CVE and no reported malware, infostealer, or supply-chain compromise for that version.
5. **Pick the best candidates** that meet the trust bar. Add them following the project's package manager workflow and pinning policy.

## Fast Checks With gh

The GitHub CLI answers the star and advisory questions faster than a web search. Use it first, then use `/research-web` for compromises too recent or too broad for the advisory database, such as a registry or maintainer takeover still being reported.

Stars, activity, and archive status:

```sh
gh repo view OWNER/REPO --json nameWithOwner,stargazerCount,pushedAt,isArchived
```

Advisories for the exact planned version, from the GitHub Advisory Database. `ECOSYSTEM` is one of `npm`, `pip`, `rust`, `go`, `maven`, `nuget`, `rubygems`, `composer`, `pub`, `swift`, `erlang`, or `actions`:

```sh
gh api -X GET /advisories -f ecosystem=ECOSYSTEM -f affects=PACKAGE@VERSION \
  --jq '.[] | [.ghsa_id, .cve_id, .severity, .published_at, .summary] | @tsv'
```

Malware is excluded from that query by default, so always run it a second time with `-f type=malware`, without a version to catch any malicious release of the package:

```sh
gh api -X GET /advisories -f ecosystem=ECOSYSTEM -f affects=PACKAGE -f type=malware \
  --jq '.[] | [.ghsa_id, .published_at, .summary] | @tsv'
```

Advisories the project published itself:

```sh
gh api repos/OWNER/REPO/security-advisories --jq '.[] | [.ghsa_id, .cve_id, .severity, .state, .summary] | @tsv'
```

An empty result is not proof of safety: the database covers these ecosystems only, and new reports land with a delay.

Prefer a stable version published at least three weeks ago over a fresh release to reduce supply-chain attack risk, unless the work needs a feature or fix from a newer version.

## Approval

- Add a candidate that meets the trust bar without asking.
- Ask for explicit approval for every candidate that does not, including any whose stars or security status cannot be verified. Say which part of the bar it misses.
- Group all candidates that need approval into one question, or a few, instead of asking one by one.
- Skip the approval request for dependencies that are already a settled decision, such as those named in an approved plan. Still run step 3 on them, since a compromise may have been reported since the decision.

## Stop On Malware

🛑 If a dependency that is about to be added, or already was added, fetched, or installed, has known malware, an infostealer, or a supply-chain compromise:

1. Stop immediately and do not continue the task under any circumstances.
2. Do not install, build, run, or import anything that could load it.
3. Check, read-only, whether it was already fetched, installed, or executed, and where: manifests, lockfiles, package caches, the Nix store, virtual environments, and shell history.
4. Tell the user under a 🛑 heading what was found, which parts are affected, and the report that identified it. Leave cleanup to the user.

## Output

For each dependency considered, report briefly:

- The candidate and version, and whether it was added, needs approval, or was rejected.
- Its GitHub stars and the security check result, with sources for anything found.
- For candidates that need approval, which part of the trust bar they miss.
Pliki w paczce (2)
  • SKILL.md 5 KB
  • agents/openai.yaml 1 KB

Wersje

W bazie nie ma innych skilli o nazwie dependency-add.

Migawka bazy z 13.11.2026 · wersja paczki f8403994c822 (skrót SHA-256 całej paczki)

Pobrania w ostatnich 30 dniach: 1 · ranking

Zgłoś problem z tym skillem: szkodliwa treść, błędny opis albo prośba autora o usunięcie.

Podobne skille

  • 100dependency-tree-compress
    Szuka małych, mało zaufanych zależności w drzewie projektu, które można zastąpić własnym kodem i zmniejsza ryzyko.
  • 100dependency-hygiene
    Pomaga utrzymać zależności małe, aktualne i bezpieczne: decyzje o dodaniu pakietu, audyt drzewa, konfiguracja Dependabota lub Renovate, konflikty wersji i usuwanie porzuconych bibliotek.
  • 90dependencies
    Odpowiada na pytania o zadeklarowane zależności projektu, wersje bibliotek i ograniczenia pakietów, gdy ich źródła nie są w kontekście.
  • 100architecture-model-sync
    Aktualizuje model architektury, gdy dodano lub usunięto crate, kontekst ograniczony, zależność zewnętrzną albo relację między nimi.
  • 90dependency-update-review
    Przegląda aktualizacje zależności, zmiany menedżera pakietów i diffy lockfile, oceniając ryzyko przed wdrożeniem.
  • 100shared-packages
    Ustala wzorce wspólnych pakietów monorepo dla szybkich MVP: dodawanie bibliotek, narzędzi i importów między pakietami.

Uwagi i zgłoszenia

Wiadomość trafi do Bartosza Sokolińskiego. Adres e-mail jest potrzebny tylko do odpowiedzi. Szczegóły w polityce prywatności.