jig-ui 0.14.0 → 0.15.0

package/CHANGELOG.md

@@ -1,5 +1,72 @@
 # Changelog
 
+## 0.15.0
+
+The gate holds an agent on warnings, waits while it asks the owner a question,
+and reaches projects where the skill is installed globally.
+
+### Changed
+
+- **`gate` blocks on warnings in the files the agent changed**, not only on
+  mechanical errors. One em dash survived two runs of fresh agents on a real
+  site: the first saw the warning, called it pre-existing, and finished. The
+  gate still reads only changed files, so old warnings elsewhere never trap an
+  agent, and it now also runs when a changed file carries interface text.
+- **The skill asks the user to run each command.** `decide`, `spec`, `mockup`,
+  `make` and `critique` hold their procedures in the command, not in the
+  skill. An agent told in prose to "run decide" made up five questions of its
+  own and wrote a `DECISIONS.md` with no direction and no `## Unresolved`
+  section.
+
+### Added
+
+- **`jig-allow <ID>: <why>`** waives one warning, on its own line or the line
+  above, in any comment syntax. Some warnings are a detector's guess: `A-01`
+  cannot tell a chosen violet from a default one. Warnings only, a reason is
+  required, and `check` lists every waiver it honoured, with its reason, on
+  every run. Waived findings are left out of `warnings=`.
+- **`install --agent claude --hook` on a global install** writes only this
+  project's Stop hook. The hook was written only by a project install, which is
+  refused while the skill is global, so most projects never had a gate.
+
+### Fixed
+
+- **A question to the owner is not a finish.** After `decide`, `spec` or
+  `mockup`, a stop whose last message asks a question is allowed, and the
+  command's output is checked on the stop after the answer. Before, the gate
+  refused `decide`'s first question because `DECISIONS.md` did not exist yet,
+  and on the second refusal the agent wrote the file itself with no answers.
+- **The gate finds `DECISIONS.md` beside the token layer** that `brand` in
+  `jig.config.json` names. It looked in three fixed places, rejected a
+  correctly placed file, and the agent moved the file to satisfy it.
+- **`jig seo` says an endpoint is not a page**, so `pages=` can be checked
+  against a site's route files.
+
+## 0.14.1
+
+`jig seo` cites only what a rule says.
+
+### Fixed
+
+- **No finding for a missing sitemap or robots file.** Both were filed under
+  rules that do not ask for them: `J-123` is about a page that must not be
+  indexed and says `robots.txt` is not it, and `J-124` is about a sitemap that
+  contradicts the pages. A project with no domain yet cannot write an honest
+  sitemap, so the only way to clear the old finding was to invent one. Both are
+  now counts on the `JIG_SEO` line.
+- **A sitemap of paths is a `J-124` error.** A `<loc>` that is `/about` rather
+  than a full URL was read as a route, although a crawler drops it. `J-124` now
+  names that case and says no sitemap is the honest one until there is an
+  origin.
+- **`pages=` counts pages.** It counted any file that mentioned `<title>` or
+  `<meta>`, so a build script, a test, a data module and a layout were pages and
+  a home page inheriting its layout was not. A page is now a route a framework
+  serves, or a whole-document template outside the layout and partial folders.
+  A dynamic route counts once.
+- **Framework routes match their sitemap entries.** `src/pages/admin` was read
+  as `/pages/admin`, so a `noindex` Astro or Next page listed in the sitemap
+  went unreported.
+
 ## 0.14.0
 
 What a stranger meets before the page, what the page must not hand them, and a

package/README.md

@@ -53,10 +53,19 @@ Every agent supports both scopes.
 answer yes when an interactive install offers it. It adds one entry to
 `.claude/settings.json`, keeping everything already in that file, and runs `jig gate`
 when the agent tries to finish: it holds the agent there while the files it changed
-fail `check`, or while the `/jig` command it just ran left its work incomplete — a
-spec that is prose, a critique with no verdict files, a review that never rendered
-the page. After three attempts it lets the agent stop and says the work is
-unfinished.
+have a `check` error or warning, or while the `/jig` command it just ran left its
+work incomplete — a spec that is prose, a critique with no verdict files, a review
+that never rendered the page. After three attempts it lets the agent stop and says
+the work is unfinished.
+
+A warning that is right as it stands is waived on its own line, in a comment
+reading `jig-allow <ID>: <why>` (`/* jig-allow A-01: the brand is violet */`). The
+reason is required, an error cannot be waived, and `check` lists every waiver it
+honoured, with its reason, on every run.
+
+The skill and the hook are installed separately. With Jig installed globally, run
+`install --agent claude --hook` in a project to add only that project's hook: a
+global hook would run on every stop in every project on the machine.
 
 It is off by default, and `--yes` never adds it, because that is the path an agent
 takes and nobody is there to consent. It exists because weak models skip any step
@@ -357,7 +366,7 @@ overwrites a config or brand file you have edited.
 | `init [--yes]` | Sets the project up: CSS system, brand colour, token files, `jig.config.json`, wired imports, baseline check. The only command that writes into your repo. |
 | `check [--all] [--ci] [--json]` | Runs the rules a machine can decide. Reports findings by rule id. |
 | `update` | Refreshes an install to a newer version, leaving alone any file you have edited. |
-| `seo [--json]` | Audits what a search engine and a link preview read, across the whole project: a route whose metadata says `noindex` sitting in the sitemap, two pages claiming one title, indexable pages with no sitemap, a sitemap that lists nothing. Needs no config, no decisions and no spec. |
+| `seo [--json]` | Audits what a search engine and a link preview read, across the whole project: a route whose metadata says `noindex` sitting in the sitemap, two pages claiming one title, a sitemap that lists nothing or lists paths a crawler drops. Whether a sitemap and a robots file exist is counted, not reported: no rule asks for either, and a site with no domain yet cannot write an honest sitemap. Needs no config, no decisions and no spec. |
 | `verdicts <surface>` | Verifies a critique's verdict files and computes its counts: every rule in each pass judged once, no id that does not exist, no rule in the wrong arm, and no verdict the render probe contradicts. |
 | `probe` | Prints the render probe — one expression the critique runs in a browser at each width. It operates the menu, measures sideways scroll, and reads whether the styles and tokens applied. |
 | `gate` | Run by the Stop hook `install` adds for Claude Code, not by hand. Blocks an agent from finishing while `check` fails on the files it changed, or the step it just ran left its work unfinished. |
@@ -448,8 +457,8 @@ in it tells strangers where to look; a path is safe to name only when something
 else protects it.
 
 `jig seo` covers what one file cannot: a route that says `noindex` and sits in
-the sitemap anyway, two pages claiming one title, a site whose pages nothing
-points at. It writes nothing and needs nothing, so it is safe to run on the first
+the sitemap anyway, two pages claiming one title, a sitemap of paths a crawler
+drops. It writes nothing and needs nothing, so it is safe to run on the first
 day, or on somebody else's codebase.
 
 ## Using it with a coding agent