Skip to content

The engine renders: an aligned card, aligned tables, drawn trees, a palette, and the width to fit them in - #153

Merged
wmadden-electric merged 12 commits into
mainfrom
engine-colour
Aug 11, 2026
Merged

The engine renders: an aligned card, aligned tables, drawn trees, a palette, and the width to fit them in#153
wmadden-electric merged 12 commits into
mainfrom
engine-colour

Conversation

@wmadden-electric

@wmadden-electric wmadden-electric commented Aug 11, 2026

Copy link
Copy Markdown
Contributor

What this looks like

prisma-cli auth workspace list, before and after:

name  id  status              name      id    status
Acme Inc  ws_1  current  →    Acme Inc  ws_1  current
Globex  ws_2                  Globex    ws_2

prisma-cli auth whoami, whose card is the shape 36 commands use:

status: signed in                status:     signed in
user: bob@example.com       →    user:       bob@example.com
workspace: Acme Inc              workspace:  Acme Inc

And the tree the style guide has specified since it was written, which until now no block could express:

✘ contract
├─ ✘ table user
│  └─ ✔ primary key: id
└─ ✔ table post

Headers and keys cyan, connectors dim, glyph and label coloured by the node's status. None of that was reachable before: makeUi emitted two escape sequences, bold and dim, and nothing else in the engine emitted colour at all.

What was broken

The engine shipped a common renderer that did almost no rendering. renderBlock was columns.join(" ") — no sizing, no padding. The key/value card printed label: value, having lost the aligned, accent-coloured column the commander shell drew. The tree emitted two-space indents and had zero users. There was no way to draw a migration graph's lane gutter at all, and OutputStream was { write(text) }, so the engine could not learn the terminal's width even from the bin.

Both consumers had already lost rendering to it. The platform port flattened auth workspace list, project list and agent status, recorded as accepted divergences. The ORM's renderers ship colourless, and two of them call createColors({ useColor: true }) to force colour past NO_COLOR because there is no other way to get any.

The decision

The engine renders; a command describes. So the fix is to the renderer, not to give commands a canvas.

  • table aligns and fields draws the card again — keys padded to one column, toned, with the dim │ rail available as fields.rail for the commands that want it.
  • tree draws connectors and status glyphs, matching the style guide verbatim.
  • drawing is added as the escape hatch for layout the engine cannot derive: the migration DAG's lane gutter, where lane assignment comes from a BFS over the graph and one hue must reach the gutter cell, the node glyph and the label alike.
  • Text = string | Span[] wherever a block takes display text. Spans carry tone, never escape sequences, so the engine measures on text and colour cannot break alignment. The pad-versus-colour bug that exists in shipped code becomes unrepresentable.
  • Status is separate from Tone. One says what happened and picks the glyph; the other says what colour to paint. They were fused in summary.tone, which meant a tree node could not be a failure painted in its branch lane's colour.
  • ui.width is stderr's width, unbounded when stderr is not a terminal. Only the command knows what to sacrifice, so it is told the room it has; if it overruns anyway the engine prints it unmodified.

Two corrections ride along. Colour resolution keyed off stdout while blocks render to stderr, so cmd > file disabled colour for output a human was watching — it now follows the stream that is actually printed to. And every one of the CLI's four failure surfaces drew ✖, while the style guide and docs/product/output-conventions.md have both specified ✘ all along; all four now draw ✘.

Nothing here was invented

The palette is not a set of tastes. Every tone is the colour a shipping Prisma CLI already uses for that meaning, checked at source during review:
heading, and the eight other platform semantics packages/cli/src/shell/ui.ts
identifier, ref, highlight the ORM CLI's formatters in prisma/prisma
color-1 … color-6 the ORM's migration-graph LANE_COLORIZERS, in order

The indexed colours are six rather than the eight first sketched, because six is what the ORM's lane rotation is, and it excludes red deliberately so no lane can read as an error. Basic ANSI has no two further hues that stay distinguishable from those six; reaching eight would mean a 256-colour palette neither CLI uses. colorette and string-width are the colour and measurement libraries both repos already depend on.

How to believe the card came back

packages/cli/tests/v8-golden-rendering.test.ts renders a fields block through the engine and asserts it equals what the commander shell's own renderFieldRows produces. It is a comparison against the original code in another package, building its own colour instance and sharing nothing with the engine — not a byte string someone typed into a test.

Alignment is proven not to depend on colour: a table whose cells carry spans is byte-identical to the same table in plain text once escapes are stripped, which a colour-first-then-pad implementation would fail. A CJK cell is measured in terminal columns rather than code units.

What this does not do

No command binds ui yet — all 54 platform presenters are still written human: () => [...], so the engine colours what it draws itself and nothing more. Spans, drawing and ui.width get their first callers when each family converts its renderers, which is the next step for both. Next actions and diagnostics still render as plain strings, so half of stderr is coloured and half is not; that is the natural follow-on slice and it is recorded as such.

The rail is available but no command sets it. Alignment and colour return to all 36 cards at once; which of them want the rail is a per-command judgement the engine cannot make.

Alternatives considered

A separate graphic presenter alongside human — rejected: a drawing is human output, so it belongs in the human channel's grammar, and a parallel presenter would need its own channel and colour handling.

Routing rich renderings through Presentations.stdout — rejected: that is the machine channel, and a drawing's consumer is a human.

Letting commands draw tables and trees themselves — rejected: if every command reimplements every form of rendering on an open canvas, the common renderer has no purpose.

Keeping NO_COLOR absolute over an explicit --color — rejected: a flag on the invocation beats a standing environment default. NO_COLOR still beats the terminal.

Supersedes #148, which carried this spec alone. The spec is in this PR at .drive/projects/prisma-cli-v8/specs/engine-colour.md, with the four amendments the implementation survey produced recorded in its §8.

🤖 Generated with Claude Code

wmadden-electric and others added 12 commits August 11, 2026 18:08
@wmadden-electric @claude
The contract for engine-owned rendering: palette, cards, tables, tree…
5ab89c1 
…s, width

The ruled spec plus its dispatch plan. Four amendments came out of reading
what prisma/prisma and prisma/composer already ship, rather than deciding
by taste: six indexed colours instead of eight (the ORM's lane rotation is
six and deliberately excludes red, and basic ANSI has no two further hues
that stay distinguishable); status split out of tone, so a tree node can be
a failure painted in its branch lane's colour; the fields card restored,
because it is the most-used block and lost the most; and an explicit
--color beating NO_COLOR.

Every tone's rendering is taken from a shipping implementation, and the
colour and width libraries are the ones both repos already depend on.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Signed-off-by: willbot <w.a.madden+machine@gmail.com>
Signed-off-by: Will Madden <madden@prisma.io>
@wmadden-electric @claude
The engine can paint a tone, and knows how wide stderr is
a008b9c 
The engine shipped a common renderer that emitted exactly two escape
sequences, bold and dim, and could not reach the terminal's width at
all: OutputStream was write-only, structurally, even from the bin.

Ui now carries tone(), which turns one of the contract's tones into
bytes, and width, which is stderr's columns — unbounded when stderr is
not a terminal, so the arithmetic a renderer does needs no special case.
Every rendering in the palette is taken from a shipping CLI rather than
chosen fresh: the platform shell's semantic colours and the ORM CLI's
formatters, whose six-colour lane rotation becomes color-1..color-6 and
keeps its exclusion of red. Colour off makes every verb the identity
function, so a renderer has one code path.

Text, Span, Tone and Status are exported for the block grammar to take
up next; no block renders differently yet.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Signed-off-by: willbot <w.a.madden+machine@gmail.com>
Signed-off-by: Will Madden <madden@prisma.io>
@wmadden-electric @claude
Colour follows the stream the engine actually prints to
c360efa 
Blocks, diagnostics and commentary all render to stderr, but the colour
decision was keyed on stdout. So `cmd > file` stripped colour from
output a human was still watching on the terminal, and `cmd 2> file`
left escape sequences in a file nobody reads.

Colour now follows stderr, and the precedence is stated explicitly:
--color or --no-color decides whatever else is set, otherwise NO_COLOR
disables it, otherwise stderr's being a terminal decides. An explicit
flag on the invocation beats anything in the environment.

sniffFormat and resolveAutoFormat still read stdout, which is right:
format asks whether stdout is a pipe, which is a different question.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Signed-off-by: willbot <w.a.madden+machine@gmail.com>
Signed-off-by: Will Madden <madden@prisma.io>
@wmadden-electric @claude
A terminal resized mid-run reports its new width
ac4fbee 
The bin copied process.stderr.columns once, when it assembled the
runtime, so ui.width was frozen at whatever the terminal was when the
process started. Node updates that property on SIGWINCH, so reading it
through a getter costs nothing and means a session command drawing
minutes into a run sees the width the terminal has now.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Signed-off-by: willbot <w.a.madden+machine@gmail.com>
Signed-off-by: Will Madden <madden@prisma.io>
@wmadden-electric @claude
A summary states what happened, not what colour to be
12d6a20 
summary.tone was doing two jobs: it picked the glyph AND was the only
colour word a block had. Fusing them means a block cannot say "this
failed" and "paint it in this hue" at once — which the tree needs, so
a failing node can be drawn in its branch lane's colour.

The field becomes summary.status, typed Status, and selects the glyph
alone. An optional tone overrides the colour the status implies and
never touches the glyph. 57 call sites renamed with it, compiler-driven.

The error glyph moves from ✖ to ✘, the character the contract names for
Status and the one the style guide's tree example and the legacy shell
both use. No summary in the CLI is an error today, so no output changes;
a diagnostic's severity symbol is a different map and keeps its ✖.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Signed-off-by: willbot <w.a.madden+machine@gmail.com>
Signed-off-by: Will Madden <madden@prisma.io>
@wmadden-electric @claude
The engine draws the blocks it was only listing before
d294636 
renderBlock was five one-liners. A table was columns.join("  "), so
nothing lined up; a card was `label: value`, which is the alignment and
the accent colour the platform port recorded as an accepted loss; a tree
was two-space indents with no connectors at all.

Every block text field now takes Text — a string or spans — and the
engine turns it into bytes. A table sizes each column to its widest
cell, measured on the text rather than the escape sequences, so a
coloured cell aligns exactly as the same words uncoloured; the last
column is never padded, so no line trails whitespace. A card pads
`label:` to a common width with the label in the accent colour, and
draws the dim rail when a command asks for one. A tree draws ├─, └─ and
the │ continuation in dim, with the status glyph before the label.

A new drawing block writes lines of spans verbatim, for output whose
two-dimensional structure the engine cannot derive — a migration graph's
lane gutter, where one hue has to reach the gutter cell, the node glyph
and the label alike.

The card and rail are asserted against the commander shell's own
renderFieldRows and renderVerboseBlock rather than a copied byte string,
so "the card came back" is checked, not claimed.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Signed-off-by: willbot <w.a.madden+machine@gmail.com>
Signed-off-by: Will Madden <madden@prisma.io>
@wmadden-electric @claude
One mark for a failure, on every surface that reports one
cf9ce9f 
Three maps in the engine pick the character that marks a failure: a
block's status, a step's outcome, and a diagnostic's severity. D2 moved
the first to ✘, the character the style guide, the contract and the
legacy shell all name, and left the other two on ✖. A run could report a
failed step with one mark, the summary with another, and the diagnostic
explaining it with a third.

All three now draw ✘, and the internal-error note in reporting.ts with
them. ✔, ⚠ and ℹ were already one character each across the maps; two
tests hold all of it in place by running one command that reports a
failure through every surface at once, and one that does the same for
every non-failure outcome.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Signed-off-by: willbot <w.a.madden+machine@gmail.com>
Signed-off-by: Will Madden <madden@prisma.io>
@wmadden-electric @claude
The colour probe describes its text instead of pre-rendering it
7a769f3 
The command the colour-resolution tests run spliced ui.tone(...) into a
summary's text string. That is the pattern Text and Span exist to
replace: a handler that builds bytes is a handler whose width the engine
cannot measure. The probe now hands over a span and lets the engine
paint it, so the test demonstrates the path commands are meant to take
rather than the one being retired.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Signed-off-by: willbot <w.a.madden+machine@gmail.com>
Signed-off-by: Will Madden <madden@prisma.io>
@wmadden-electric @claude
ui.width is checked to be read on every access, not once
9345e1c 
The contract requires the width be read per render rather than cached,
and nothing covered it — the field went from a frozen snapshot to a
getter inside this branch and the suite was green either way.

The test harness cannot express this: it reads the caller's columns once
while assembling the runtime, so the value can never change under a
running command. The check is therefore a direct one against makeUi with
a stub stream whose columns the test mutates, which is also what lets it
cover the Ui's other verbs — that emphasize and dim still emit the exact
bytes they emitted before colorette replaced the hand-rolled sequences.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Signed-off-by: willbot <w.a.madden+machine@gmail.com>
Signed-off-by: Will Madden <madden@prisma.io>
@wmadden-electric @claude
The bin's width passthrough is checked to be live, not copied
c4751b0 
Two getters carry the terminal's width to a command: the Ui's, and the
bin's forwarding of process.stderr.columns. Only the first was covered,
and the second is the one that regressed inside this branch — it shipped
as a copy taken while assembling the runtime, and was replaced by a
getter, with the suite green both times.

The engine-side stub cannot see past itself, so a copy reintroduced in
the bin would freeze ui.width at process start with every existing test
still passing. This asserts against the bin adapter directly: the fake
process moves its stderr width after assembleRuntime has returned, and
the runtime reports the new value. Reverting the getter to a copy fails
it.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Signed-off-by: willbot <w.a.madden+machine@gmail.com>
Signed-off-by: Will Madden <madden@prisma.io>
@wmadden-electric @claude
The internal-error note's mark is pinned along with its text
a5d5306 
reporting.ts writes its mark as a literal instead of reading the
diagnostic-severity map, so the map-level pins do not reach it. The only
test that renders that line matched the message text and stopped short
of the mark, which left the one error glyph in the engine that nothing
held in place. The assertion now covers the whole line — mark, code and
message. Changing the mark in reporting.ts fails it.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Signed-off-by: willbot <w.a.madden+machine@gmail.com>
Signed-off-by: Will Madden <madden@prisma.io>
@wmadden-electric @claude
Say what the divergence lists and the interface draft now describe
420b799 
Three parity entries recorded the flattened rendering as accepted. The
card's alignment and its accent colour are back and byte-equal to the
legacy renderer, so those entries now say what returned and what did
not — the rail, which is a per-command opt-in, and the header framing,
which has no engine counterpart. The interface draft's presentation
section matches what shipped, since the exports barrel cites it as
normative.

Two files argued for deferring glyph mode on the grounds that the engine
already emits a character set it no longer draws; the failure mark is
now the style guide's throughout, so the premise is corrected and the
deferral stands on the true one.

Also recorded: a test that reports green against a stale engine build
when run as a lone pnpm filter, and a second spawn test that fails under
load.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Signed-off-by: willbot <w.a.madden+machine@gmail.com>
Signed-off-by: Will Madden <madden@prisma.io>
@coderabbitai

coderabbitai Bot commented Aug 11, 2026 •
edited
Loading

Copy link
Copy Markdown

Warning

Review limit reached

@wmadden, you've reached your PR review limit, so we couldn't start this review.

Next review available in: 51 minutes

You've used all free OSS reviews for now. Wait for the free limit to reset to keep reviewing this public repository.

How can I continue?

After more reviews become available, a review can be triggered using the @coderabbitai review command as a PR comment. Alternatively, push new commits to this PR.

To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews.

How do review limits work?

CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability.

For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window.

Please refer docs for additional details.

Review details
⚙️ Run configuration

Configuration used: Organization UI

Review profile: ASSERTIVE

Plan: Pro Plus

Run ID: afc6b6f3-d8f0-4410-8fa1-6d7130d91fe5

📥 Commits

Reviewing files that changed from the base of the PR and between 8c43266 and 420b799.

⛔ Files ignored due to path filters (1)
  • pnpm-lock.yaml is excluded by !**/pnpm-lock.yaml
📒 Files selected for processing (86)
  • .drive/projects/prisma-cli-v8/assets/engine/engine-interface-draft.ts
  • .drive/projects/prisma-cli-v8/assets/engine/whoami-parity-divergences.md
  • .drive/projects/prisma-cli-v8/assets/s2/parity-divergences-s2b.md
  • .drive/projects/prisma-cli-v8/assets/s2/parity-divergences-s2c.md
  • .drive/projects/prisma-cli-v8/deferred.md
  • .drive/projects/prisma-cli-v8/plans/engine-colour.md
  • .drive/projects/prisma-cli-v8/specs/engine-colour.md
  • .drive/projects/prisma-cli-v8/specs/s2b-design/facts/facts-v8-patterns.md
  • packages/cli-engine/package.json
  • packages/cli-engine/src/execution/command-context.ts
  • packages/cli-engine/src/execution/palette.ts
  • packages/cli-engine/src/execution/rendering.ts
  • packages/cli-engine/src/execution/reporting.ts
  • packages/cli-engine/src/execution/shared-flags.ts
  • packages/cli-engine/src/exports/index.ts
  • packages/cli-engine/src/presentation.ts
  • packages/cli-engine/src/runtime.ts
  • packages/cli-engine/src/testing.ts
  • packages/cli-engine/tests/blocks.test.ts
  • packages/cli-engine/tests/clack-isolation.test.ts
  • packages/cli-engine/tests/clack-prompts.test.ts
  • packages/cli-engine/tests/command-families.test.ts
  • packages/cli-engine/tests/config.test.ts
  • packages/cli-engine/tests/events.test.ts
  • packages/cli-engine/tests/execution.test.ts
  • packages/cli-engine/tests/interaction-affordances.test.ts
  • packages/cli-engine/tests/palette.test.ts
  • packages/cli-engine/tests/prompts.test.ts
  • packages/cli-engine/tests/redirects.test.ts
  • packages/cli-engine/tests/run-hooks.test.ts
  • packages/cli/src/v8/agent/presentation.ts
  • packages/cli/src/v8/auth/login.ts
  • packages/cli/src/v8/auth/logout.ts
  • packages/cli/src/v8/auth/whoami.ts
  • packages/cli/src/v8/auth/workspace-list.ts
  • packages/cli/src/v8/auth/workspace-logout.ts
  • packages/cli/src/v8/auth/workspace-use.ts
  • packages/cli/src/v8/branch/list.ts
  • packages/cli/src/v8/bucket/create.ts
  • packages/cli/src/v8/bucket/delete.ts
  • packages/cli/src/v8/bucket/key-create.ts
  • packages/cli/src/v8/bucket/key-delete.ts
  • packages/cli/src/v8/bucket/key-list.ts
  • packages/cli/src/v8/bucket/list.ts
  • packages/cli/src/v8/feedback.ts
  • packages/cli/src/v8/git/connect.ts
  • packages/cli/src/v8/git/disconnect.ts
  • packages/cli/src/v8/postgres/backup-list.ts
  • packages/cli/src/v8/postgres/connection-list.ts
  • packages/cli/src/v8/postgres/connection-remove.ts
  • packages/cli/src/v8/postgres/list.ts
  • packages/cli/src/v8/postgres/presentation.ts
  • packages/cli/src/v8/postgres/remove.ts
  • packages/cli/src/v8/postgres/restore.ts
  • packages/cli/src/v8/postgres/show.ts
  • packages/cli/src/v8/postgres/usage.ts
  • packages/cli/src/v8/project/env-add.ts
  • packages/cli/src/v8/project/env-list.ts
  • packages/cli/src/v8/project/env-remove.ts
  • packages/cli/src/v8/project/env-shared.ts
  • packages/cli/src/v8/project/env-update.ts
  • packages/cli/src/v8/project/list.ts
  • packages/cli/src/v8/project/presentation.ts
  • packages/cli/src/v8/project/remove.ts
  • packages/cli/src/v8/project/rename.ts
  • packages/cli/src/v8/project/show.ts
  • packages/cli/src/v8/project/transfer.ts
  • packages/cli/src/v8/runtime.ts
  • packages/cli/src/v8/service/presentation.ts
  • packages/cli/src/v8/telemetry/consent.ts
  • packages/cli/src/v8/telemetry/status.ts
  • packages/cli/tests/v8-auth.test.ts
  • packages/cli/tests/v8-bin.test.ts
  • packages/cli/tests/v8-branch.test.ts
  • packages/cli/tests/v8-bucket.test.ts
  • packages/cli/tests/v8-git.test.ts
  • packages/cli/tests/v8-golden-rendering.test.ts
  • packages/cli/tests/v8-postgres.test.ts
  • packages/cli/tests/v8-project.test.ts
  • packages/cli/tests/v8-service-domain.test.ts
  • packages/cli/tests/v8-service-open.test.ts
  • packages/cli/tests/v8-service-promote.test.ts
  • packages/cli/tests/v8-service-remove.test.ts
  • packages/cli/tests/v8-service-rollback.test.ts
  • packages/cli/tests/v8-service-session.test.ts
  • packages/cli/tests/v8-whoami.test.ts

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

Comment @coderabbitai help to get the list of available commands.

@wmadden-electric wmadden-electric mentioned this pull request Aug 11, 2026
Closed
@pkg-pr-new

pkg-pr-new Bot commented Aug 11, 2026

Copy link
Copy Markdown

Open in StackBlitz

npx https://pkg.pr.new/@prisma/cli@153

npx https://pkg.pr.new/@prisma/cli-engine@153

commit: 420b799

@wmadden

wmadden commented Aug 11, 2026

Copy link
Copy Markdown

@coderabbitai review this

@coderabbitai

coderabbitai Bot commented Aug 11, 2026 •
edited
Loading

Copy link
Copy Markdown

@wmadden I will review pull request #153.

⚠️ Action not completed

Review rate limited.

Note: CodeRabbit is an incremental review system and does not re-review already reviewed commits. This command is applicable only when automatic reviews are paused.

@wmadden-electric
wmadden-electric merged commit 784db69 into main Aug 11, 2026
11 checks passed
@wmadden-electric
wmadden-electric deleted the engine-colour branch August 11, 2026 19:09
wmadden-electric added a commit that referenced this pull request Aug 11, 2026
@wmadden-electric
Merge main into spec/engine-owns-telemetry
f5d1aa8 
Two modify/delete conflicts, both on files this branch removed in D4:
packages/cli/src/v8/telemetry/consent.ts and status.ts. The deletions
stand — the engine ships those commands now — but #153 edited them as
part of the rendering rework, so its change had to land on the engine's
copies or the engine's telemetry commands would render unlike every
other command in the CLI.

What #153 changed there was summary blocks: tone: became status:, since
Status (what happened, which picks the glyph) is now separate from Tone
(what colour to paint). Carried onto src/telemetry/commands.ts for both
the status summary and the enable/disable confirmation, and onto the four
test command definitions on this branch that build summary blocks.

Nothing else needed carrying, and I checked rather than assumed: the
fields card takes Text where it took string, so its plain rows still
typecheck; human: () => [...] is still assignable now that the presenter
receives a Ui; and no v8 command sets fields.rail or emits spans — the
PR says so and the tree agrees — so telemetry adopting either would have
made it the odd one out rather than consistent.

Verified by running them, not by compiling them. telemetry status draws
a blue info glyph and a cyan-keyed card whose values align in one column
— the same shapes auth whoami draws — and telemetry disable draws the
bright green tick every ok summary draws. stdout still carries only the
machine lines.

Signed-off-by: willbot <w.a.madden+machine@gmail.com>
Signed-off-by: Will Madden <madden@prisma.io>

#144 touches project commands and merged without interaction.
wmadden-electric added a commit that referenced this pull request Aug 11, 2026
@wmadden-electric @claude
test(cli-engine): match the error glyph main now renders
023ebb6 
The new findings tests byte-assert stderr, and #153 changed the error
marker from U+2716 to U+2718 while this branch was open.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Signed-off-by: willbot <w.a.madden+machine@gmail.com>
Signed-off-by: Will Madden <madden@prisma.io>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@wmadden wmadden wmadden approved these changes

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

2 participants

@wmadden-electric @wmadden