j.connect
Skill0 starsChanged 22 days agoApache-2.0View on GitHub
Guided service setup — pick a service from the descriptors on disk, detect your platform, then install its CLI, authenticate, register its MCP server and independently verify, skipping whatever is already done and never handling a secret value.
In this repository151 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- j.brainstormSkill
- j.breakSkill
- j.btwSkill
- j.check-inSkill
- j.clearifySkill
- j.close-storySkill
- j.cloud-connectSkill
- j.commitSkill
- j.continueSkill
- j.conventionsSkill
- j.dashboard-shareSkill
- j.dashboardSkill
- j.deep-diveSkill
- j.distributeSkill
- j.docSkill
- j.doc-syncSkill
- j.doooSkill
- j.doSkill
- j.jenga-configSkill
- j.jenga-permission-levelSkill
- j.jengaSkill
- j.errorSkill
- j.evaluateSkill
- j.examplifySkill
- j.gitignoreSkill
- j.helpSkill
- j.ideaSkill
- j.improveSkill
- j.initSkill
- j.jbpSkill
- j.lgtmSkill
- j.pi-planSkill
- j.playbook-newSkill
- j.playbookSkill
- j.proceedSkill
- j.publishSkill
- j.reconcile-originSkill
- j.reconcileSkill
- j.redoSkill
- j.skillifySkill
- j.smoke-testSkill
- j.spinoffSkill
- j.statusSkill
- j.testSkill
- j.todoSkill
- j.toolsSkill
- j.unchartedSkill
- j.wtfSkill
- skill-nameSkill
- j.brainstormSkill
- j.breakSkill
- j.btwSkill
- j.check-inSkill
- j.clearifySkill
- j.close-storySkill
- j.cloud-connectSkill
- j.commitSkill
- j.connectSkill
- The first 60 shown.
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 1 is distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 44
Resolve the service: If the user already named a service, resolve it against services by id or name.
- Distinctive: line 56
If the user types an id or name, resolve it against the list.
- Very specific: line 66
If status is unknown: say the platform (or a usable package manager) could not be determined and print fallback_docs_url as the official …
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Connect — Guided Service Setup
Purpose
E65_S02 is the single entry point for wiring a third-party service into a project: install its CLI,
authenticate, register its MCP server where one exists, and independently verify the result. Everything
deterministic already lives in standalone scripts under skills/j-connect/scripts/, driven by per-service
descriptors (project/documentation/service-descriptor.md). Per this repo's "Scripts Over Inline Logic" principle
(CLAUDE.md), this SKILL.md holds only judgment and presentation: it sequences four scripts, asks the
user what to do at decision points, and relays their output. It contains no install, detection,
validation, registration or reporting logic of its own, and no service names or service list.
Instructions
List the services. Invoke:
bash skills/j-connect/scripts/list-services.shIt prints one JSON object:
services(each withid,name,file,mcp,requires,unresolved_requires) andskipped(descriptors that failed validation). Relay anyskippedentries to the user so a broken descriptor is visible. Ifservicesis empty, say that no service descriptors are installed yet, point atproject/documentation/service-descriptor.mdfor how to add one, and stop.Resolve the service. If the user already named a service, resolve it against
servicesbyidorname; never guess a match. Otherwise present the picker, built only from theservicesarray in the order given, using the numbered-options format with free text as the final option:Which service would you like to connect? 1. <name of first service> 2. <name of second service> ... N. Other (type a service id)If the user types an id or name, resolve it against the list. An unknown value is not guessed at: say it is not available and show the picker again. Mention any
unresolved_requiresof the chosen service (a prerequisite descriptor that is not installed) before continuing, since the run will report it.Detect the platform. Invoke, with the chosen entry's
file:bash skills/j-connect/scripts/detect-platform.sh --descriptor <file>If
statusisunknown: say the platform (or a usable package manager) could not be determined and printfallback_docs_urlas the official instructions for the user to follow themselves, then continue to step 4. The runner skips the install step when the CLI is already present, and otherwise reports it asneeds-user-actionwith the same URL without running anything. Never guess, construct or suggest an install command.Run the setup. Invoke:
bash skills/j-connect/scripts/run-descriptor.sh <file>It prints one JSON line per step (
pass,fail,skipped,needs-user-action) and stops at the firstfailorneeds-user-action. Relay each step to the user, and say plainly which steps were skipped because they were already satisfied (for example an already installed CLI or an already authenticated account). When the install step is reportedneeds-user-actionand a usable package manager is available, ask:The CLI is not installed. How would you like to proceed? 1. Install it now (re-run with --allow-install) 2. I will install it myself from the official instructions 3. Other (describe below)Re-run with
--allow-installonly after the user explicitly picks option 1; never pass it on your own initiative.- Hand over human steps. For any other
needs-user-action(a browser sign-in, setting a token environment variable, approving the MCP server in Claude Code), relay the instruction verbatim, wait for the user to say they have done it, then re-run the samerun-descriptor.shcommand. It is idempotent and skips completed steps, so repeat until it stops asking. Secrets rules, with no exceptions: refer to environment variables by name only; never ask the user to paste a secret value, token or key into the chat; never read, print or echo the contents of a secrets or env file. If a secrets file is needed, the runner itself makes it git-ignored before creating it — do not create or edit one yourself. Report the result. Pipe the final run's output through the report script, for example by saving the run's stdout to a results file and invoking:
bash skills/j-connect/scripts/summarize-run.sh <results-file>Relay its output to the user verbatim. Claim success only when its last line starts with
VERIFIED:; that line comes from the service's own independent verification. The user saying a step is done is never evidence of success. If the last line isNOT VERIFIED, report that, quote the step it names, and offer to re-run step 4 or stop.
No session-end behaviour
This skill performs no session-end work: no handoff file, no queue write, no board or status update.
This matches j.cloud-connect and j.dashboard-share.
Out of Scope
- Per-service logic of any kind. Adding a service is a descriptor-only change (a new
<id>.jsonunderskills/j-connect/descriptors/); this file never changes for it, and it holds no service list. - Install, detection, authentication, MCP registration, git-ignore guardrails and verification — all of
that is the scripts' scope (
E65_S01,E65_S02_T01–T03). Do not duplicate or reimplement any part of it here, and do not run vendor commands directly. - App Store Connect and Play Console credentials — handled by
j.publish(E65epic scope note).
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

