jfrog-setup-package-managers

name	jfrog-setup-package-managers
description	Use this skill when the user asks to set up, configure, bind, or connect a package manager (npm, pip, uv, pipenv, maven, gradle, go, docker, helm, ...) to JFrog Artifactory via `jf setup` and `.jfrog/local/package-resolution.json`; when a workspace manifest exists with no matching binding entry; or when a session hook reports package-manager config missing. Skip when the binding already has the same repo key. Never pick a repo by discovery; use resolver output only (unless the user names or asks to browse repos). On unresolved or failed setup, ask with the failure verbatim — never switch servers.
metadata	{"role":"workflow"}

JFrog — Setup Package Managers for Artifactory

Apply the session hook's repo pick via jf setup, then record it in .jfrog/local/package-resolution.json. jf setup writes package-manager-native config (.npmrc, pip.conf, uv.toml, …); the binding lets the hook re-apply on later sessions.

Scope (this skill vs session hook)

Session-start hook: resolves repo keys per package type, injects the "Resolved URLs for this session" table, refreshes the global cache. The same renderer is available on demand via modules/package-resolution/scripts/print-policy.mjs (the enforce notice embeds the exact command), so the policy can be loaded after setup.

This skill: reads that output, runs jf setup, and persists the workspace binding at .jfrog/local/package-resolution.json when package-manager config is still missing.

Honor the injected policy's governed scope. The session policy lists the package managers it governs. Do not proactively onboard a package manager the policy doesn't govern (e.g. a stray Dockerfile when only pypi/npm are governed) — those are intentionally out of scope. An explicit user request to set up any package manager still works (Step 1's user-mention signal and Step 2's AskQuestion for an unlisted package manager apply as usual).

Prerequisites

• jf setup mutates user state (~/.npmrc, ~/.docker/config.json, …). Confirm before the first jf setup in a session unless the user explicitly requests silent/non-interactive setup.
• Reading ../jfrog/SKILL.md is required — done as Step 0.1 below.

Out of scope: CLI install/login (../jfrog/references/…).

Gotchas

• Always pass --repo and --server-id — omitting --repo fails when multiple repos match. See jf-setup-command.md.
• jf setup overwrites package-manager config without backup — skip package managers whose binding already matches (Step 1, signal 2).
• Docker / Podman — prefix or stop. jf setup docker writes creds only; bare docker pull <img> hits Docker Hub. Complete setup, then pull via <host>/<repoKey>/<img>.
• Binding holds decisions, not credentials — never write tokens into .jfrog/local/package-resolution.json.
• gradle ≠ maven. Bind under repositories.gradle, never repositories.maven.
• Yarn / Poetry — not APR zero-touch; bind only on explicit user ask (Step 1).

References

File	When to read
references/jf-setup-command.md	CLI flags, supported package managers, exit-code contract, jf setup --help
references/global-cache-file.md	Global cache shape, resolution classes, jq one-liners
references/workspace-binding.md	Workspace binding schema, package-manager → type map, merge semantics

Step 0 — Read the base skill, then ensure jf is ready

1. Read ../jfrog/SKILL.md fully first — always, before any jf command, even when jf is already configured. It carries the jf invariants this skill relies on. After reading, run that skill's Environment check (and export JFROG_CLI_USER_AGENT) before the first jf call.
2. Ensure jf + a configured server (<SID>). If jf config show already succeeds, skip to Step 1; otherwise:
   • jf --version missing → install per ../jfrog/references/jfrog-cli-install-upgrade.md.
   • jf config show empty → login per ../jfrog/references/jfrog-login-flow.md or jf config add with access-token (Bearer-only).
3. Do not run jf setup until both succeed. Confirm before install/login.

Step 1 — Identify package managers to bind

Combine four signals, in order; intersect with jf setup --help supported list:

1. Explicit user mention. Map aliases: python → pip/uv/pipenv (and poetry only if the user named Poetry); java → maven/gradle; node → npm/pnpm by lockfile (yarn only if the user named Yarn).

2. Workspace binding — read .jfrog/local/package-resolution.json. Drop package managers already bound to the same key unless recovering from 401/403 (re-run same key). Package-manager → type table: workspace-binding.md.

3. Workspace manifests when still ambiguous (several package managers of one type may apply — e.g. requirements.txt and uv.lock):

   Manifest / signal	Package manager
   package.json, pnpm-lock.yaml	npm (+ pnpm if pnpm-lock.yaml present)
   yarn.lock (alone)	npm — do not auto-select yarn
   requirements.txt	pip
   Pipfile	pipenv
   uv.lock	uv — suppresses bare pyproject.toml → pip; keep requirements.txt + uv.lock as multi-PM
   pyproject.toml	[tool.uv] → uv; [tool.poetry] → poetry only on explicit user ask, else not applicable (do not select pip); bare PEP 621 with no uv.lock →

Step 2 — Get the resolved repo

For each <package-manager>, recover <repoKey> and <serverId> from the first source available:

1. "Resolved URLs for this session" table (default). Parse <repoKey> from URL; <serverId> from host.
2. Workspace binding — if table was trimmed. repositories.<type> (gradle → repositories.gradle, not maven).
3. Global cache — last resort only; never overrides (1) or (2). See global-cache-file.md.

Cache disagreeing with (1)/(2) is not a reason to change the repo.

Don't choose a repo yourself: no listing, enumerating, probing, or iterating --server-id to pick one, and don't second-guess the resolver — use resolver output only. If the user explicitly asks to browse repos, list them via jf api "/artifactory/api/repositories?type=virtual&packageType=<pkgType>" (Artifactory package type from the binding map — gradle not maven; uv / pip / pipenv / poetry → pypi), then let the user choose; the agent still never makes the choice on its own.

Unresolved repo key

Ask via AskQuestion (include the resolver/setup failure text verbatim):

No default repo for <package-manager> on <SID>. Failure: <verbatim failure> Which Artifactory repository should I use? (repo key, or abort.)

Cap at 2 answers per package manager, then abort. User may override repo only, never server.

Step 3 — Confirm, run jf setup, persist binding

1. Present the plan, one row per package manager:

   <package-manager>  → <repoKey> on <SID>               (source: resolver)
   <package-manager>  → <repoKey> on <SID>               (source: user-supplied)
   

2. Show binding diffs when the repo key changes.

3. Confirm via AskQuestion (apply / change repos / abort) unless the user explicitly requested silent/non-interactive setup — then run directly.

4. Sequentially, one package manager at a time:

   jf setup <package-manager> --server-id <SID> --repo <repoKey> [--project <key>]
   

5. Exit code 0 = success — merge binding (step 6). On non-zero, stop, surface CLI output verbatim, offer alternate repo or abort (2-answer cap).

6. On success, merge into .jfrog/local/package-resolution.json per workspace-binding.md:

   { "repositories": { "<pkgType>": "<repoKey>" } }
   

   Map package manager → type via the reference table (gradle → gradle). Merge atomically.

Step 4 — Load the routing policy

If this session started with the "routing NOT READY" (enforce) notice, that notice includes a refresh command (node <plugin>/modules/package-resolution/scripts/print-policy.mjs). After Step 3 succeeds, run that exact command and treat its stdout as the authoritative, now-current policy — it prints the resolved Artifactory URLs and hard rules. Continue the original request using those URLs.

If the command prints nothing, routing is off by config (packageResolution.enabled is not true) — an admin opt-in. Report that to the user and let them decide whether to enable it.