Skip to content

feat: consume coding-agent-kit v0.3.0 and preview account details in fzf-tab - #5

Merged
FridrichMethod merged 3 commits into
mainfrom
feat/account-preview
Oct 21, 2026
Merged

FridrichMethod merged 3 commits into
mainfrom
feat/account-preview

Conversation

@FridrichMethod

@FridrichMethod FridrichMethod commented Oct 21, 2026 •

Copy link
Copy Markdown
Owner

What

Consumes coding-agent-kit v0.3.0 (FridrichMethod/coding-agent-kit#2, tag v0.3.0 on 081f54a) and shows account details in the fzf-tab completion preview. It builds on #4 (merged), the kit's own Stow package.

  1. chore(deps): update coding-agent-kit submodule to v0.3.0.
    • The gitlink moves to the v0.3.0 tag. A new minor release is refused by the daily bump, so this reviewed PR is the path.
    • Every test entry point and tests/run.ps1 also unset CODING_AGENT_KIT_USAGE_FILE, CODING_AGENT_KIT_TOKEN_ID and XDG_STATE_HOME. The delivery knob gate requires the two kit names.
  2. feat(zsh): preview account details in fzf-tab completion.
    • custom/fzf-tab.zsh previews a codex-as or claude-as account with --account-info NAME, offline and without opening a credential.
    • It branches on fzf-tab's $group ([Codex accounts], [Claude accounts]), because the launchers' completions have no per-tag context.
    • Any other word, or a name that is no account, previews as itself.
    • A credential path still never reaches the generic file preview.
  3. fix(tests): import pty only in the real fzf-tab case.
    • The first CI run failed only the native Windows job: a module-level import pty needs termios, which Windows Python lacks, so test_ai_integration could not be imported there.
    • The import now sits inside the one case that needs it, and that case is skipped on Windows.
    • CI is green on all three OSes. The 7-host bootstrap-e2e ran on commit 2 and passed; commit 3 changes only that test import.

Verified (lab clones, push disabled, temporary homes)

  • tests/run.sh --ci, the pwsh suites and pre-commit run --all-files pass on each commit alone. e2e-harness-windows fails only in the lab, from its long TMPDIR, as on refactor(coding-agent-kit): give the kit its own Stow package #4.
  • Preview tests:
    • fail on the old zstyle and pass on the new one;
    • one is hermetic, on the context fzf-tab really looks up;
    • one drives real fzf-tab in an interactive zsh with a fake fzf. It needs FZF_TAB_TEST_DIR and skips without it.
  • Parity against refactor(coding-agent-kit): give the kit its own Stow package #4, 6 Unix hosts x Stow 2.4.1/2.3.1: path sets, link texts and realpaths are equal. The only differences are the contents of the changed kit files and fzf-tab.zsh, plus the two new --help lines.

Not verified here

  • Native Windows: CI and the e2e-ci/** run are the gate.
  • CI has no fzf-tab clone, so the real-fzf-tab case skips there.

A new minor release, so a reviewed bump rather than the daily patch run.
v0.3.0 keeps last-seen account details offline and adds --account-info
and --account-labels; claude-as now passes its child
CODING_AGENT_KIT_USAGE_FILE and CODING_AGENT_KIT_TOKEN_ID. Every test
entry point and tests/run.ps1 unset those two and XDG_STATE_HOME, so a
caller's session never reaches a suite (the delivery knob gate requires
the two kit names).
The fzf-tab preview of codex-as and claude-as now shows the kit's offline
--account-info for an account (last-seen email or label, plan and usage)
and the word itself for any other argument, so a credential path still
never reaches the generic file preview. fzf-tab looks this context up by
command only (the completions call _describe without _arguments), so the
group, the [%d] description, tells an account from an argument; a failing
query falls back to the word.

The old test evaluated the zstyle under a context fzf-tab never uses. The
new hermetic case uses the real one, and a second case drives real
fzf-tab (FZF_TAB_TEST_DIR) in an interactive zsh with a fake fzf that runs
fzf-tab's own preview command per candidate; both fail against the old
zstyle. Docs describe the account details and the privacy trade-off.
pty imports termios, which native Windows Python lacks, so importing
test_ai_integration failed there and the Windows CI job lost the whole
module. The one case that needs pty is skipped on Windows and now imports
it itself.
@FridrichMethod
FridrichMethod marked this pull request as ready for review October 21, 2026 13:51
Copilot AI balanced review requested due to automatic review settings October 21, 2026 13:51
@FridrichMethod
FridrichMethod merged commit 974d44d into main Oct 21, 2026
3 checks passed

Copilot AI left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

@FridrichMethod
FridrichMethod deleted the feat/account-preview branch October 21, 2026 13:51
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants