Refresh Claude marketplace cache and report setup failures honestly by jeremy · Pull Request #555 · basecamp/basecamp-cli

jeremy · 2026-07-22T20:11:27Z

Fixes #417.

Root cause

basecamp setup detects Claude Code, installs the agent skill, then shells out to claude plugin install basecamp@37signals. It first runs claude plugin marketplace add, but add no-ops on an already-registered marketplace — it does not refresh the cached metadata.

The transport itself was fixed upstream: basecamp/claude-plugins#7 (merged Jul 20) switched the basecamp plugin entry from the SSH github shorthand to an anonymous-HTTPS url source. But a user who registered the marketplace before that change keeps a stale cached entry declaring source: github, so claude plugin install still clones git@github.com:basecamp/basecamp-cli.git and dies with Permission denied (publickey) on machines without a GitHub SSH key.

Reproduced against a clean Claude home:

Scenario	`marketplace add` (what setup runs)	`plugin install`
Clean home	caches `source: url`	succeeds over HTTPS ✓
Stale cache (`source: github`)	"already on disk" — no refresh	SSH clone → fails w/o key ✗
Stale cache + `marketplace update`	refreshes to `source: url`	succeeds over HTTPS ✓

Separately, the wizard printed "Setup complete!" even when the plugin step failed, hiding the failure behind a single easy-to-miss ✗.

Changes

Refresh the marketplace cache before install. Add a best-effort claude plugin marketplace update <name> after each marketplace add on the install paths (mirroring what the Codex path already does), so a stale source: github entry is replaced with the current HTTPS source: url first. Claude's own marketplace clone already falls back SSH→HTTPS, so the refresh works without a key.
Honest completion reporting. wizardAgents now returns an explicit outcome that distinguishes a deliberate skip (still "complete") from a real failure ("incomplete"). When post-setup agent checks fail, showSuccess drops "Setup complete!" for an honest headline, lists each failing step with its own agent-specific remediation hint (Claude vs Codex), and points to basecamp doctor. Exit stays 0 (graceful degrade). The machine-output summary reflects the same status.

No CLI command/flag surface changes.

Tests

Marketplace refresh asserted on both the reinstall and fresh-install non-interactive paths (via the existing stub-claude harness).
Pure-helper unit tests for collectAgentIssues, statusFromOutcome, successHeadline, and the incomplete wizardSummaryLine branch — a declined setup stays complete; a real failure becomes incomplete.

bin/ci green.

Summary by cubic

Refreshes the Claude marketplace cache before plugin installs and reports setup results honestly with a bounded refresh and clear skip behavior. This avoids SSH clone failures for basecamp@37signals, catches leftover stale plugins, and gives actionable next steps.

Bug Fixes
- Run claude plugin marketplace update 37signals after marketplace add before every plugin install (fresh installs and reinstalls, interactive and non-interactive).
- Avoids stale source: github entries that forced SSH clones; installs now use HTTPS as intended.
- Bound the marketplace refresh to 60s so a stalled clone can’t hang setup; install still proceeds if it times out.
- Re-check surviving stale plugin entries and mark the run incomplete if cleanup fails.
- Manual recovery hint now includes the marketplace update step so copy-paste fixes don’t reproduce the SSH failure.
- No CLI surface changes.
New Features
- Setup outcome is issue‑authoritative and based on a single post‑setup snapshot, so the headline and checklist never disagree.
- If the user skips agent setup, the summary shows “Coding agent setup skipped” instead of red checks; a skip remains “complete”.
- When issues remain, the banner reflects it and shows agent‑specific hints; it also points to basecamp doctor. Machine output mirrors the status and exit code stays 0.

^{Written for commit cfd575f. Summary will update on new commits.}

Fixes #417. Root cause: `basecamp setup` runs `claude plugin marketplace add` before installing the plugin, but `add` no-ops on an already-registered marketplace. Users who added the 37signals marketplace before its source metadata switched from the SSH `github` shorthand to an HTTPS `url` source (basecamp/claude-plugins#7) are left with a stale cache that makes `claude plugin install` clone via `git@github.com:` — which fails on machines without a GitHub SSH key. Separately, the wizard printed "Setup complete!" even when the plugin step failed. Refresh the cache with `claude plugin marketplace update <name>` before each install (mirroring the Codex path), so a stale `source: github` entry is replaced with the current HTTPS `source: url` first. Claude's own marketplace clone already falls back SSH→HTTPS, so the refresh works without a key. Report honestly: wizardAgents now returns an explicit outcome distinguishing a deliberate skip from a real failure. When agent checks fail after setup, the summary drops "Setup complete!" for an honest headline, lists each failing step with its own agent-specific remediation hint, and points to `basecamp doctor`. A user who declines agent setup still reads as complete.

Copilot

Pull request overview

This PR fixes setup failures when installing the Claude Code plugin from a stale marketplace cache (issue #417) by refreshing marketplace metadata before install, and it improves the setup wizard’s reporting so it no longer claims “Setup complete!” when agent/plugin setup steps actually failed.

Changes:

Refresh Claude marketplace metadata via claude plugin marketplace update 37signals (best-effort) before plugin installs, including reinstall and fresh-install non-interactive paths.
Make agent-setup return an explicit outcome (skipped vs issues) and propagate that into wizard status (complete vs incomplete) and the human/machine summaries.
Add unit tests for the new pure helpers and assertions that the marketplace refresh is invoked in relevant paths.

Tip

If you aren't ready for review, convert to a draft PR.
Click "Convert to draft" or run gh pr ready --undo.
Click "Ready for review" or run gh pr ready to reengage.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 1 comment.

File	Description
internal/commands/wizard.go	Makes the completion headline and summary reflect incomplete agent setup, and prints per-issue remediation hints.
internal/commands/wizard_agents.go	Adds marketplace refresh before Claude plugin install and returns structured agent-setup outcomes/issues for honest reporting.
internal/commands/wizard_test.go	Adds coverage for the new helper logic and asserts Claude marketplace refresh is invoked in non-interactive paths.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

cubic-dev-ai

All reported issues were addressed across 3 files

_{Reply with feedback, questions, or to request a fix.

Re-trigger cubic}

…tatus Address review follow-ups on the #417 fix: - Manual recovery hint now includes `claude plugin marketplace update`, so the copy-paste path also refreshes a stale cache instead of reproducing the SSH clone failure (marketplace add alone no-ops on an existing marketplace). - Completion status and the rendered checklist now derive from a single post-setup snapshot (snapshotAgentChecks) carried through the outcome, so a transient check can no longer make the headline and checklist disagree. - Status is issue-authoritative: any observed failure marks the run incomplete and Skipped no longer suppresses one. A deliberate skip records no issues, so it stays complete. - Tests: rendered showSuccess assertions (no "Setup complete!" on failure, agent-specific hint present, doctor pointer present); exact add → update → install ordering on both the non-interactive and interactive install paths; snapshot/issue helpers covered with supplied agents (no global registry mutation).

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 7aa4588818

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

cubic-dev-ai

All reported issues were addressed across 3 files (changes from recent commits).

_{Reply with feedback, questions, or to request a fix.

Re-trigger cubic}

Copilot

Pull request overview

Copilot reviewed 3 out of 3 changed files in this pull request and generated 1 comment.

Comments suppressed due to low confidence (1)

internal/commands/wizard.go:371

The remediation list prefixes the agent name even though the check names already include it (e.g. “Codex — Codex Plugin”), which results in redundant output. Consider using the check name as-is here to keep the remediation lines concise.

			label := issue.Check
			if issue.Agent != "" {
				label = issue.Agent + " — " + issue.Check
			}

…ip UX Address the next review round: - Bound the marketplace refresh with a 60s child context so a stalled clone can't hang setup; install still proceeds on timeout. - Re-check stale plugin keys after handlers (claudeStaleIssues): a current- plugin health check can pass while stale entries survive removal, which previously still reported complete. Leftover stale keys now mark the run incomplete. - Skip renders coherently: a declined agent setup shows "Coding agent setup skipped" instead of red failing checks under a "complete" headline. - Remediation no longer double-names agents whose check already carries the name ("Claude Code Plugin", not "Claude Code — Claude Code Plugin"). - Doc/comment accuracy: describe issues as snapshot failures plus standalone setup failures; drop a stale step number. - Tests: rendered skip-path summary; stale post-condition → incomplete; exact add → update → install ordering on the reinstall path too.

Copilot

Pull request overview

Copilot reviewed 3 out of 3 changed files in this pull request and generated no new comments.

Copilot AI review requested due to automatic review settings July 22, 2026 20:11

github-actions Bot added commands CLI command implementations tests Tests (unit and e2e) labels Jul 22, 2026

Copilot started reviewing on behalf of jeremy July 22, 2026 20:11 View session

github-actions Bot added the bug Something isn't working label Jul 22, 2026

github-advanced-security AI found potential problems Jul 22, 2026

View reviewed changes

Comment thread internal/commands/wizard_agents.go

Copilot AI reviewed Jul 22, 2026

View reviewed changes

Comment thread internal/commands/wizard.go

cubic-dev-ai Bot reviewed Jul 22, 2026

View reviewed changes

Comment thread internal/commands/wizard_agents.go

Copilot AI review requested due to automatic review settings July 22, 2026 20:53

Copilot started reviewing on behalf of jeremy July 22, 2026 20:53 View session

chatgpt-codex-connector Bot reviewed Jul 22, 2026

View reviewed changes

Comment thread internal/commands/wizard_agents.go

cubic-dev-ai Bot reviewed Jul 22, 2026

View reviewed changes

Comment thread internal/commands/wizard.go Outdated

Copilot AI reviewed Jul 22, 2026

View reviewed changes

Comment thread internal/commands/wizard.go Outdated

Copilot AI review requested due to automatic review settings July 22, 2026 21:23

Copilot started reviewing on behalf of jeremy July 22, 2026 21:23 View session

Copilot AI reviewed Jul 22, 2026

View reviewed changes

jeremy merged commit eb2f5e5 into main Jul 22, 2026
26 of 27 checks passed

jeremy deleted the fix/plugin-install-https-fallback branch July 22, 2026 22:44

jeremy mentioned this pull request Jul 22, 2026

Produce gosec SARIF via golangci-lint for suppression parity #561

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Refresh Claude marketplace cache and report setup failures honestly#555

Refresh Claude marketplace cache and report setup failures honestly#555
jeremy merged 3 commits into
mainfrom
fix/plugin-install-https-fallback

jeremy commented Jul 22, 2026 •

edited by cubic-dev-ai Bot

Loading

Uh oh!

Uh oh!

Copilot AI left a comment

Uh oh!

Uh oh!

cubic-dev-ai Bot left a comment •

edited

Loading

Uh oh!

Uh oh!

chatgpt-codex-connector Bot left a comment

Uh oh!

Uh oh!

cubic-dev-ai Bot left a comment •

edited

Loading

Uh oh!

Uh oh!

Copilot AI left a comment

Uh oh!

Uh oh!

Copilot AI left a comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Uh oh!

Conversation

jeremy commented Jul 22, 2026 • edited by cubic-dev-ai Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Root cause

Changes

Tests

Summary by cubic

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Uh oh!

cubic-dev-ai Bot left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

cubic-dev-ai Bot left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

jeremy commented Jul 22, 2026 •

edited by cubic-dev-ai Bot

Loading

cubic-dev-ai Bot left a comment •

edited

Loading

cubic-dev-ai Bot left a comment •

edited

Loading