The design/pages/review prompt
The whole text of config/agent-prompts/design/pages/review.md, as an agent is handed it.
Bundled agent prompts — one row of 30, exported from
config/agent-prompts
· every prompt
· the resolution order
What this prompt is
| id | design/pages/review | the key every tier, every pattern step and every override spells it by |
| kind | stage instruction | whether any bundled pattern hands it to an agent step |
| role | review | the harness role the agent reading it runs under |
| stage | review | the stage of a run it is handed at |
| bundled at | config/agent-prompts/design/pages/review.md | where the built-in copy sits in Gantry's own tree |
Where an override of it goes
| tier | the file to write | what it is |
|---|---|---|
|
run
consulted 1st |
.gantry/<plan>/prompts/design/pages/review.md |
the run's own frozen copy, written when the run started and read by every agent it launches; edit it to change what a run's remaining agents are told |
|
project
consulted 2nd |
.gantry/<plan>/design/pages/review.md |
a per-run override an operator drops beside the run's other orchestration files |
|
global
consulted 3rd |
~/.config/gantry/agent-prompts/design/pages/review.md |
the materialized global mirror; it overrides the built-in once its bytes differ from the built-in's |
|
built-in
consulted 4th |
config/agent-prompts/design/pages/review.md |
the default compiled into the binary |
The text
Review the built page against its design brief and this project's design system. The gate proves only
that tests pass and the route renders; judge whether the page is good design for this app.
Check that it composes the existing system, uses tokens rather than stray literal styling, honours the
brief's signature, serves every required function, handles every required state on mobile and desktop,
and reads as native to the app. Improve safe issues directly. If it needs another pass, write the review
scope verdict and retry note requested by the standard review contract. If it changes assumptions for
remaining tickets, express that by editing the scoped ticket files the active pattern permits.
When you route, use the run history in your environment section. Repeated failure on one ticket is
evidence about the ticket: repeat the same ticket only when the prior note named one specific gap,
this attempt closed it, and the rest is plainly reachable in one more pass. Judge the whole ticket
each time, not just what the previous review complained about - attempts that each deliver a
different fragment mean the ticket asks for too many independent things; say so and edit the scoped
ticket files when permitted. The attempt count is never a reason to lower the bar: do not mark a final
attempt `complete` to let the run proceed - the attempt budget exists so an unconvergeable ticket
stops visibly and resumably.
Write `{{VERDICT_PATH}}` as TOML holding exactly one of `value = "complete"`, `value = "retry"`, or
`value = "blocked"`. The file holds that one TOML assignment and nothing else - no prose verdict line.
When choosing `retry`, write the one-sentence reason to `{{RETRY_NOTE_PATH}}`.
Brief:
{{BRIEF}}
Diff:
{{DIFF}}