@skitterbyte/skitterspec-linear 19.16.0 → 19.18.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/MIGRATION.md +78 -0
- package/README.md +0 -1
- package/assets/claude-md-section.md +6 -9
- package/assets/commands/allow-main.md +1 -2
- package/assets/core/env.config.json.example +0 -1
- package/assets/core/env.config.md +24 -27
- package/assets/review/page.html +10 -83
- package/assets/rules/spec-planning.md +14 -21
- package/assets/rules/spec-reports.md +112 -54
- package/assets/skills/no-spec/SKILL.md +2 -2
- package/assets/skills/spec/SKILL.md +2 -2
- package/assets/skills/spec-bug/SKILL.md +10 -13
- package/assets/skills/spec-diff/SKILL.md +47 -115
- package/assets/skills/spec-hotfix/SKILL.md +10 -13
- package/assets/skills/spec-next/SKILL.md +59 -42
- package/assets/skills/spec-review/SKILL.md +2 -2
- package/assets/skills/spec-reviewed/SKILL.md +1 -8
- package/package.json +1 -1
- package/src/cli.js +251 -204
- package/src/env/config.js +16 -7
- package/src/env/hooks.js +6 -3
- package/src/env/live.js +48 -0
- package/src/env/review.js +71 -173
- package/src/env/serve.js +3 -7
- package/src/init.js +509 -13
- package/assets/commands/spec-remote-review.md +0 -37
package/MIGRATION.md
CHANGED
|
@@ -1,5 +1,83 @@
|
|
|
1
1
|
# Migration guide
|
|
2
2
|
|
|
3
|
+
## A PR-mode finish runs the `afterLand` entries its land matched (nothing to edit)
|
|
4
|
+
|
|
5
|
+
After the host merged a PR, `spec-env integrate` used to print `nothing to
|
|
6
|
+
run` and list every conditional `afterLand` entry under `skipped (nothing in
|
|
7
|
+
the land matched)`, whatever its `when` said. The range it read was the
|
|
8
|
+
branch's own (`origin/main...branch`), and a merged branch is already inside
|
|
9
|
+
`origin/main`, so that range was always empty.
|
|
10
|
+
|
|
11
|
+
The finish now reads what the primary checkout's base gains by catching up
|
|
12
|
+
(`main...origin/main`), so an entry whose glob matches a file in the land
|
|
13
|
+
runs, with `# <path> changed`. Where base was already caught up by hand,
|
|
14
|
+
nothing lies between the two refs. That range is treated as unknown, and the
|
|
15
|
+
conditional entries run with `# range unknown`.
|
|
16
|
+
|
|
17
|
+
Your globs were right: a leading `**/` already matched a top-level path. If
|
|
18
|
+
a PR-mode land skipped an install or a generate, run it by hand in the
|
|
19
|
+
primary checkout once. Later lands run it for you.
|
|
20
|
+
|
|
21
|
+
## A local install stays local in a worktree (check for stray team files)
|
|
22
|
+
|
|
23
|
+
`update` run inside a worktree of a local install (`init --local`) used to run
|
|
24
|
+
as a **team** install. The manifest that records `"scope": "local"` is kept out
|
|
25
|
+
of git, and `spec-env bootstrap` did not seed it, so the worktree had none.
|
|
26
|
+
That run then:
|
|
27
|
+
|
|
28
|
+
- appended the `<!-- skitterspec:start -->` block to the **tracked**
|
|
29
|
+
`CLAUDE.md`,
|
|
30
|
+
- created an untracked `.claude/settings.json` holding skitterspec's hooks,
|
|
31
|
+
- listed unedited managed files as `customized (kept)`,
|
|
32
|
+
- and wrote the worktree a manifest with no `scope`.
|
|
33
|
+
|
|
34
|
+
Bootstrap now seeds the manifest with the other local-install files. `update`
|
|
35
|
+
also reads a local scope from the `.git/info/exclude` block when a worktree
|
|
36
|
+
has no manifest of its own, so a worktree bootstrapped before this release is
|
|
37
|
+
covered without re-bootstrapping. It reports
|
|
38
|
+
`local install (recorded in .git/info/exclude) — kept local`.
|
|
39
|
+
|
|
40
|
+
**If an earlier `update` already ran in such a worktree, revert it by hand.**
|
|
41
|
+
Nothing removes it for you, because a `CLAUDE.md` edit or a
|
|
42
|
+
`.claude/settings.json` might be yours:
|
|
43
|
+
|
|
44
|
+
```
|
|
45
|
+
git checkout -- CLAUDE.md # drop the appended block
|
|
46
|
+
rm .claude/settings.json # only if skitterspec created it
|
|
47
|
+
rm specs/.core/.skitterspec-manifest.json # the scopeless one it wrote
|
|
48
|
+
skitterspec spec-env bootstrap <spec> # re-seeds the primary's manifest
|
|
49
|
+
```
|
|
50
|
+
|
|
51
|
+
Run these in the worktree. A worktree that never ran `update` needs nothing.
|
|
52
|
+
|
|
53
|
+
|
|
54
|
+
## The tailnet is `remote`; publishing the review is retired (nothing to edit)
|
|
55
|
+
|
|
56
|
+
The review stack's `remote:` line used to be a claude.ai page the review was
|
|
57
|
+
published to. That tier is gone. `remote:` is now the review server's
|
|
58
|
+
**tailnet address**, on its own line, labelled `(tailnet)`. It is the same
|
|
59
|
+
server as `local` and `network`, so one wait covers all three, and a verdict
|
|
60
|
+
pressed there wakes the waiting session like any other.
|
|
61
|
+
|
|
62
|
+
```
|
|
63
|
+
local: http://127.0.0.1:7760/7e9e…/feat-orders
|
|
64
|
+
network: http://192.168.0.136:7760/7e9e…/feat-orders
|
|
65
|
+
remote: http://100.126.111.34:7760/7e9e…/feat-orders (tailnet)
|
|
66
|
+
```
|
|
67
|
+
|
|
68
|
+
- **`/spec-remote-review` is removed.** `skitterspec update` retires the command
|
|
69
|
+
file. A copy you edited is kept, with the usual warning, and you can delete
|
|
70
|
+
it.
|
|
71
|
+
- **`review.allowRemote` is ignored.** A config that still sets it loads
|
|
72
|
+
without a warning, and you can delete the key whenever you like.
|
|
73
|
+
`spec-env review allow remote` now refuses and names the tailnet line.
|
|
74
|
+
- **`remote` follows `review.allowNetwork`.** It is off whenever `network` is,
|
|
75
|
+
with the same command to turn it on. On a machine with no tailnet address it
|
|
76
|
+
reads `— (no tailnet address on this machine)`.
|
|
77
|
+
- **`--publish-copy` is gone**, along with the page's published mode and its
|
|
78
|
+
`Allow remote` button. A page published before this change still exists, and
|
|
79
|
+
`spec-env down` still names it at teardown.
|
|
80
|
+
|
|
3
81
|
## `issue` and `incident` spec kinds, per-type states, incident log (additive — nothing breaks)
|
|
4
82
|
|
|
5
83
|
Specs can now be **issues** (`issue-` prefix — non-code work done as recorded
|
package/README.md
CHANGED
|
@@ -96,7 +96,6 @@ them. Your `specs/` content and live `.core` config are never touched.
|
|
|
96
96
|
| `/spec-init` | Bootstrap or repair the workflow in a project |
|
|
97
97
|
| `/spec-connect` | Point your canonical `localhost` ports at one spec's stack |
|
|
98
98
|
| `/spec-live` | Check one spec out in the primary checkout, so your running server reloads it |
|
|
99
|
-
| `/spec-remote-review` | Permit (or forbid) reviewing from off your network |
|
|
100
99
|
| `/allow-main` | Allow writes on the base branch for this session — yours to run, never Claude's |
|
|
101
100
|
| `/allow-project` | Let this session work on another skitterspec project — yours to run, never Claude's |
|
|
102
101
|
| `/spec-skip` | Move on from a phase without a verdict, reason recorded — yours to run, never Claude's |
|
|
@@ -12,8 +12,8 @@ skill that wants one will tell you to type it.
|
|
|
12
12
|
|
|
13
13
|
**Seeing the work.** A phase is built in its own worktree, so `git diff` in your
|
|
14
14
|
terminal answers about the base branch. **`/spec-diff`** renders that worktree's
|
|
15
|
-
diff as a self-contained HTML page you open locally — or
|
|
16
|
-
|
|
15
|
+
diff as a self-contained HTML page you open locally — or on a phone, over your
|
|
16
|
+
LAN or tailnet. The diff never passes through the model, so the page costs no context
|
|
17
17
|
tokens however large it is; the optional written review is the part that costs,
|
|
18
18
|
and it is offered rather than assumed. `/spec-next` writes the page at the end of
|
|
19
19
|
every phase. **`skitterspec status`** lists what is in flight across worktrees.
|
|
@@ -98,13 +98,10 @@ running and concluded nothing, so the phase still owes an answer.
|
|
|
98
98
|
|
|
99
99
|
**Every render lists where the page can be read**, labelled, in a fixed order —
|
|
100
100
|
`local`, `network`, `remote` — each either a URL or the one command that turns it
|
|
101
|
-
on, plus a `live:` line saying whether the change is also running.
|
|
102
|
-
|
|
103
|
-
|
|
104
|
-
|
|
105
|
-
**`/spec-remote-review`**, which toggles it. Publishing is permitted by that
|
|
106
|
-
command, never performed by it: a published page is one this tooling cannot
|
|
107
|
-
remove.
|
|
101
|
+
on, plus a `live:` line saying whether the change is also running. They are
|
|
102
|
+
labelled by address: `local` is this machine, `network` your LAN, `remote` your
|
|
103
|
+
tailnet. All three are doors into one room — the same server and the same
|
|
104
|
+
waiting verdict — so one wait covers every one the stack shows.
|
|
108
105
|
|
|
109
106
|
**One ending.** Every spec skill finishes with the same block — a verdict
|
|
110
107
|
(`✅` · `⚠️` · `❌` · `⏸`), then a table of the fields that skill declares,
|
|
@@ -26,8 +26,7 @@ id it degrades to a repo-wide allow that outlives the session and is cleared onl
|
|
|
26
26
|
by `/allow-main off`; the engine says which kind you got, so do not read a bare
|
|
27
27
|
"allowed" as temporary.
|
|
28
28
|
|
|
29
|
-
**It is machine-local and gitignored.**
|
|
30
|
-
nothing anyone else pulls: it is your decision about your checkout, not the
|
|
29
|
+
**It is machine-local and gitignored.** It writes nothing anyone else pulls: it is your decision about your checkout, not the
|
|
31
30
|
project's policy. The project-wide switch is `guards.mainIsLandingZone` in
|
|
32
31
|
`specs/.core/env.config.json`, and turning that off is a different, committed
|
|
33
32
|
decision.
|
|
@@ -202,10 +202,13 @@ through to a documented conservative default; see the field notes below.
|
|
|
202
202
|
// The changed files are what the land brings into the base branch:
|
|
203
203
|
// `base...branch` for `integrate` (read before the rebase, so it is the
|
|
204
204
|
// branch's own changes), and the carried fix commits — minus the release's
|
|
205
|
-
// version commit — for `hotfix land`.
|
|
205
|
+
// version commit — for `hotfix land`. A PR-mode finish (the host has
|
|
206
|
+
// merged) reads `base...<remote>/<base>`: what the primary checkout's base
|
|
207
|
+
// gains by catching up, since the landed branch's own range is empty.
|
|
206
208
|
// When that cannot be known, the conditional entries RUN. That is the
|
|
207
209
|
// "already landed" plan (the land happened elsewhere, so its range is not
|
|
208
|
-
// here)
|
|
210
|
+
// here), a PR-mode finish whose base was already caught up by hand, and
|
|
211
|
+
// any diff git would not answer. A redundant generate costs
|
|
209
212
|
// seconds; a skipped one leaves the base red with nothing pointing at why.
|
|
210
213
|
// A land that changed no files is known, not unknown, and skips them.
|
|
211
214
|
// The plan says why for every entry: a matched one ends in
|
|
@@ -601,14 +604,13 @@ through to a documented conservative default; see the field notes below.
|
|
|
601
604
|
// It replaced `serveOnRemote`, which gated serving on the reader — and so
|
|
602
605
|
// handed a local machine a page it could read and not answer. A legacy
|
|
603
606
|
// `serveOnRemote: false` is still read as `serve: "never"`.
|
|
604
|
-
// Either way NOTHING IS PUBLISHED
|
|
605
|
-
//
|
|
606
|
-
// remove, so that half stays an explicit ask. Teardown names a server that
|
|
607
|
+
// Either way NOTHING IS PUBLISHED: a server is one process ended by one flag,
|
|
608
|
+
// and nothing in this workflow publishes the page. Teardown names a server that
|
|
607
609
|
// served the last spec, and `spec-env prune` reaps a pidfile whose process is
|
|
608
610
|
// gone. Default: true.
|
|
609
611
|
//
|
|
610
|
-
// `allowNetwork`
|
|
611
|
-
//
|
|
612
|
+
// `allowNetwork` decides WHICH TIERS a render offers beyond `local`, and it
|
|
613
|
+
// replaced the engine guessing where the reader was sitting. It guessed
|
|
612
614
|
// for a while and got it wrong three separate ways in one day: a file:// page
|
|
613
615
|
// on a session detected `unknown`, a LAN URL for a phone that had left the
|
|
614
616
|
// network, and an address that changed underneath a reader mid-session. So
|
|
@@ -616,26 +618,21 @@ through to a documented conservative default; see the field notes below.
|
|
|
616
618
|
// turns it on — and you pick the one that reaches you.
|
|
617
619
|
//
|
|
618
620
|
// `allowNetwork` — whether the review server binds EVERY INTERFACE, so the
|
|
619
|
-
// page opens on your phone
|
|
620
|
-
//
|
|
621
|
-
//
|
|
622
|
-
//
|
|
623
|
-
//
|
|
624
|
-
//
|
|
625
|
-
//
|
|
626
|
-
//
|
|
627
|
-
//
|
|
628
|
-
//
|
|
629
|
-
//
|
|
630
|
-
//
|
|
631
|
-
//
|
|
632
|
-
//
|
|
633
|
-
//
|
|
634
|
-
// pressed on a published page needs `/spec-reviewed`, because nothing pushes
|
|
635
|
-
// from an artifact's store into a conversation.
|
|
636
|
-
//
|
|
637
|
-
// Both are toggled by `spec-env review allow <tier> --set [on|off]`, where an
|
|
638
|
-
// empty value toggles — which is what `/spec-remote-review` runs. It writes
|
|
621
|
+
// page opens on your phone (`network`, the LAN URL) and
|
|
622
|
+
// over your tailnet (`remote`), or loopback only. THIS IS
|
|
623
|
+
// WHAT CHOOSES THE BIND; `reader` no longer does.
|
|
624
|
+
// Default: true, which matches what the engine already did.
|
|
625
|
+
// `allowRemote` — RETIRED, and ignored if present. It permitted publishing
|
|
626
|
+
// the page as a claude.ai artifact; that tier is gone, and
|
|
627
|
+
// `remote` now names the tailnet address on
|
|
628
|
+
// `allowNetwork`'s switch (feat-tailnet-remote-tier).
|
|
629
|
+
//
|
|
630
|
+
// `local`, `network` and `remote` are three doors into ONE ROOM — the page
|
|
631
|
+
// POSTs to `location.pathname`, so each reaches the same server and the same
|
|
632
|
+
// waiting verdict, and one wait covers them all.
|
|
633
|
+
//
|
|
634
|
+
// It is toggled by `spec-env review allow network --set [on|off]`, where an
|
|
635
|
+
// empty value toggles. It writes
|
|
639
636
|
// THIS FILE in the primary checkout, so it changes for everyone who pulls and
|
|
640
637
|
// leaves that tree dirty; the engine says so when it does.
|
|
641
638
|
//
|
package/assets/review/page.html
CHANGED
|
@@ -2039,8 +2039,8 @@ a.ticket:hover { border-color: var(--accent); }
|
|
|
2039
2039
|
* SAID a session claimed the pass, which is the only ending with provably
|
|
2040
2040
|
* nothing left on the page to read: no code to quote, no command to run, no
|
|
2041
2041
|
* blob to paste. Every other one — still asking, established-nobody-did, a
|
|
2042
|
-
* poll that failed, a `file://` page whose pass is on the clipboard
|
|
2043
|
-
*
|
|
2042
|
+
* poll that failed, a `file://` page whose pass is on the clipboard — keeps
|
|
2043
|
+
* the tab,
|
|
2044
2044
|
* because each leaves something the reader still needs.
|
|
2045
2045
|
*
|
|
2046
2046
|
* BEST-EFFORT BY CONSTRUCTION, not by caution. Browsers refuse
|
|
@@ -2770,10 +2770,9 @@ a.ticket:hover { border-color: var(--accent); }
|
|
|
2770
2770
|
* would be worse than asking for the paste. Clear it again and they return.
|
|
2771
2771
|
* ======================================================================== */
|
|
2772
2772
|
|
|
2773
|
-
// Only a `file://` page. A served page posts
|
|
2774
|
-
//
|
|
2775
|
-
|
|
2776
|
-
var NO_TRANSPORT = location.protocol === 'file:' && !isArtifact()
|
|
2773
|
+
// Only a `file://` page. A served page posts, so it keeps its buttons —
|
|
2774
|
+
// decided from what the page IS, exactly as `send` decides.
|
|
2775
|
+
var NO_TRANSPORT = location.protocol === 'file:'
|
|
2777
2776
|
var cmdList = document.getElementById('cmd-list')
|
|
2778
2777
|
var cmdLead = document.getElementById('cmd-lead')
|
|
2779
2778
|
var cmdRows = {}
|
|
@@ -2916,85 +2915,15 @@ a.ticket:hover { border-color: var(--accent); }
|
|
|
2916
2915
|
ask()
|
|
2917
2916
|
}
|
|
2918
2917
|
|
|
2919
|
-
/**
|
|
2920
|
-
* Is this page running as a published Artifact?
|
|
2921
|
-
*
|
|
2922
|
-
* A CAPABILITY CHECK, not a hostname one. `window.claude.use` is the whole
|
|
2923
|
-
* contract for reaching anything the viewer grants, and it exists only where
|
|
2924
|
-
* a viewer is there to grant it — so its presence is what distinguishes the
|
|
2925
|
-
* published page from the two the engine serves, without this file knowing
|
|
2926
|
-
* anything about where it was published.
|
|
2927
|
-
*/
|
|
2928
|
-
function isArtifact() {
|
|
2929
|
-
return Boolean(window.claude && typeof window.claude.use === 'function')
|
|
2930
|
-
}
|
|
2931
|
-
|
|
2932
|
-
/**
|
|
2933
|
-
* Hand the pass to the artifact's own store.
|
|
2934
|
-
*
|
|
2935
|
-
* THE THIRD TRANSPORT, for the reader no local server can reach: a phone on
|
|
2936
|
-
* mobile data, or anyone away from the LAN the engine serves on. The pass
|
|
2937
|
-
* lands in a document Claude reads back and deletes, so it is consumed
|
|
2938
|
-
* exactly like a claimed one.
|
|
2939
|
-
*
|
|
2940
|
-
* `use` is asked only once the reader has pressed something. Asking at load
|
|
2941
|
-
* would spend a consent prompt on every render, including the ones nobody
|
|
2942
|
-
* ends in a verdict.
|
|
2943
|
-
*/
|
|
2944
|
-
function sendToStore(json, verdict) {
|
|
2945
|
-
copyHint.hidden = false
|
|
2946
|
-
copyHint.textContent = 'Sending…'
|
|
2947
|
-
var landed = false
|
|
2948
|
-
window.claude.use('db').then(function (db) {
|
|
2949
|
-
// `null` means this view cannot run db — not granted, not served, failed
|
|
2950
|
-
// to load, all indistinguishable by design. The clipboard is still there,
|
|
2951
|
-
// so the pass is recoverable rather than lost.
|
|
2952
|
-
if (!db) {
|
|
2953
|
-
showFallback(json, 'Could not reach the store. Copy this, then paste it to Claude.')
|
|
2954
|
-
return
|
|
2955
|
-
}
|
|
2956
|
-
return db
|
|
2957
|
-
.collection('passes')
|
|
2958
|
-
.add({ spec: data.spec, at: new Date().toISOString(), render: data.generatedAt, blob: JSON.parse(json) })
|
|
2959
|
-
.then(function () {
|
|
2960
|
-
landed = true
|
|
2961
|
-
// NO CODE HERE, and none invented. Six digits are the engine's, minted
|
|
2962
|
-
// by the holding area this pass never went through — so the page says
|
|
2963
|
-
// the one thing that is true and actionable.
|
|
2964
|
-
copyHint.textContent = 'Sent. Run this where Claude is:'
|
|
2965
|
-
showCommand('/spec-reviewed')
|
|
2966
|
-
// A PUBLISHED PAGE CANNOT ASK. The pass went into the artifact's own
|
|
2967
|
-
// store, which the engine only reads when someone tells it to, so
|
|
2968
|
-
// nothing here could ever come back `true` — and the command is the
|
|
2969
|
-
// whole ending rather than a fallback within it.
|
|
2970
|
-
markDecided(verdict, '/spec-reviewed', null)
|
|
2971
|
-
})
|
|
2972
|
-
}).then(null, function (err) {
|
|
2973
|
-
if (landed) return
|
|
2974
|
-
showFallback(
|
|
2975
|
-
json,
|
|
2976
|
-
'Could not store the pass' + (err && err.code ? ' (' + err.code + ')' : '') +
|
|
2977
|
-
'. Copy this, then paste it to Claude.',
|
|
2978
|
-
)
|
|
2979
|
-
})
|
|
2980
|
-
}
|
|
2981
|
-
|
|
2982
2918
|
function send(verdict) {
|
|
2983
2919
|
var json = JSON.stringify(buildBlob(verdict), null, 2)
|
|
2984
|
-
// DECIDED FROM WHAT THE PAGE IS, not from what it can guess. A
|
|
2985
|
-
//
|
|
2986
|
-
//
|
|
2987
|
-
// to talk to and never will.
|
|
2920
|
+
// DECIDED FROM WHAT THE PAGE IS, not from what it can guess. A page opened
|
|
2921
|
+
// over http was served by the engine and can hand the pass straight back;
|
|
2922
|
+
// a `file://` page has no server to talk to and never will.
|
|
2988
2923
|
//
|
|
2989
2924
|
// NOT "try the POST and fall back on failure": a failed POST and an absent
|
|
2990
2925
|
// one look identical to the reader, so the fallback would hide a server
|
|
2991
|
-
// that is refusing passes behind a message about copying.
|
|
2992
|
-
// is exactly that trap — its POST would go to claude.ai and fail — which is
|
|
2993
|
-
// why it is answered first, by what the page IS rather than by what failed.
|
|
2994
|
-
if (isArtifact()) {
|
|
2995
|
-
sendToStore(json, verdict)
|
|
2996
|
-
return
|
|
2997
|
-
}
|
|
2926
|
+
// that is refusing passes behind a message about copying.
|
|
2998
2927
|
if (location.protocol !== 'file:') {
|
|
2999
2928
|
post(json, verdict)
|
|
3000
2929
|
return
|
|
@@ -3037,14 +2966,12 @@ a.ticket:hover { border-color: var(--accent); }
|
|
|
3037
2966
|
// precondition of the thing they asked for, not a verdict on the diff.
|
|
3038
2967
|
'live-on': '▶ Commit & put it live',
|
|
3039
2968
|
'allow-network': '▶ Allow network',
|
|
3040
|
-
'allow-remote': '▶ Allow remote',
|
|
3041
2969
|
}
|
|
3042
2970
|
var ACTION_CMD = {
|
|
3043
2971
|
'live-on': '/spec-live',
|
|
3044
2972
|
'allow-network': 'skitterspec spec-env review allow network',
|
|
3045
|
-
'allow-remote': '/spec-remote-review',
|
|
3046
2973
|
}
|
|
3047
|
-
var SURFACE_LABEL = { live: 'live', network: 'network'
|
|
2974
|
+
var SURFACE_LABEL = { live: 'live', network: 'network' }
|
|
3048
2975
|
|
|
3049
2976
|
function surfaceStateText(row) {
|
|
3050
2977
|
if (row.kind === 'live') {
|
|
@@ -1,7 +1,7 @@
|
|
|
1
1
|
# Spec Planning
|
|
2
2
|
|
|
3
3
|
Spec-driven development is driven by ten lifecycle skills (plus the
|
|
4
|
-
`/spec-connect
|
|
4
|
+
`/spec-connect` and `/spec-live` **commands** when isolation is on) — use them rather than hand-rolling specs so the structure
|
|
5
5
|
and lifecycle stay consistent. Each sets a status on the spec header
|
|
6
6
|
(`> **Status:** …`):
|
|
7
7
|
|
|
@@ -28,7 +28,7 @@ and lifecycle stay consistent. Each sets a status on the spec header
|
|
|
28
28
|
| `/spec-init` | Bootstrap/repair this workflow in a project (idempotent) | — | — |
|
|
29
29
|
|
|
30
30
|
**Skills vs commands.** The table above lists **skills** — Claude reads them and
|
|
31
|
-
exercises judgment. `/spec-connect
|
|
31
|
+
exercises judgment. `/spec-connect` and `/spec-live` are
|
|
32
32
|
instead **slash commands**
|
|
33
33
|
(`.claude/commands/`): each pre-executes one `skitterspec spec-env` verb and
|
|
34
34
|
relays its output, so there is no judgment to apply and no model turn spent
|
|
@@ -36,19 +36,12 @@ finding one. They are marked `disable-model-invocation`, meaning
|
|
|
36
36
|
**only you can run them** — a skill that wants one will tell you to type it
|
|
37
37
|
rather than invoking it.
|
|
38
38
|
|
|
39
|
-
|
|
40
|
-
|
|
41
|
-
|
|
42
|
-
|
|
43
|
-
|
|
44
|
-
|
|
45
|
-
It exists because a render's `remote:` line has to name something a person can
|
|
46
|
-
act on, and `skitterspec spec-env review allow remote` is not what anyone
|
|
47
|
-
reconstructs from a page they are reading on a phone. Making the line
|
|
48
|
-
*clickable* was considered and rejected — the only clickable thing markdown has
|
|
49
|
-
is a URL, and a URL that acts when it is **fetched** is one a link previewer or
|
|
50
|
-
a prefetcher fires with nobody involved. It **permits** publishing; it publishes
|
|
51
|
-
nothing.
|
|
39
|
+
**The `spec-remote-review` command was retired** with the published review tier
|
|
40
|
+
(feat-tailnet-remote-tier). It toggled whether a spec might publish its review
|
|
41
|
+
as a claude.ai page — a second store no wait could watch, and a page this
|
|
42
|
+
tooling could never delete. The `remote:` line now names the review server's
|
|
43
|
+
tailnet address, on the same switch as `network`, so there is nothing left for
|
|
44
|
+
a command to toggle.
|
|
52
45
|
|
|
53
46
|
`/spec-reviewed` is a **skill** and **user-only**, and there the marking is not
|
|
54
47
|
convenience — it is the enforcement of `/spec-diff` step 0's rule that a waiting
|
|
@@ -121,8 +114,8 @@ already use, so one category is not spelled two ways. `/spec-init` asks for
|
|
|
121
114
|
both — on an existing config through a `/no-spec` branch, since the base
|
|
122
115
|
branch is a landing zone.
|
|
123
116
|
|
|
124
|
-
**`/allow-main` is the
|
|
125
|
-
from the other
|
|
117
|
+
**`/allow-main` is the third command.** It is user-only for a different reason
|
|
118
|
+
from the other two: `/spec-connect` and `/spec-live`
|
|
126
119
|
are marked that way because there is no judgment to apply. This one is marked that
|
|
127
120
|
way because **Claude must not be able to lift a guard aimed at Claude.**
|
|
128
121
|
|
|
@@ -172,7 +165,7 @@ it fails open on every cannot-tell. `/allow-project <name>` lifts it for one
|
|
|
172
165
|
session, and it is user-only for `/allow-main`'s reason: Claude must not be
|
|
173
166
|
able to lift a guard aimed at Claude.
|
|
174
167
|
|
|
175
|
-
**`/spec-skip` is the
|
|
168
|
+
**`/spec-skip` is the fourth command**, and it is user-only for `/allow-main`'s
|
|
176
169
|
reason rather than the mild one: it lifts the review gate, which is a guard
|
|
177
170
|
aimed at Claude. It pre-executes `skitterspec spec-env review skip "<reason>"`
|
|
178
171
|
and relays it — the same exit that always existed, at an address a person will
|
|
@@ -189,7 +182,7 @@ skip — never a flag. Putting the flag on `/commit` was the other suggestion an
|
|
|
189
182
|
is worse: `/commit` belongs to skittership, which is why the gate is enforced
|
|
190
183
|
by a hook rather than by editing that skill.
|
|
191
184
|
|
|
192
|
-
**`/spec-epic` is the
|
|
185
|
+
**`/spec-epic` is the fifth command**, and it is user-only for the mild reason: it
|
|
193
186
|
pre-executes `skitterspec epic status` and relays it, so there is no judgement
|
|
194
187
|
to apply. It exists for the tab an epic is run from — each member is built in
|
|
195
188
|
a tab of its own, and on a phone the in-app browser has to close to switch
|
|
@@ -380,7 +373,7 @@ the base branch — and a 350-line diff read as terminal text is scrolling, not
|
|
|
380
373
|
review. The engine collects the worktree's changes with `git -C` and writes a
|
|
381
374
|
self-contained HTML page (`.spec-env/reviews/<spec>.html`, gitignored) with
|
|
382
375
|
whole-file context that folds away, a file tree, and new files included. You open
|
|
383
|
-
it locally, or
|
|
376
|
+
it locally, or on a phone over your LAN or tailnet.
|
|
384
377
|
**The diff never passes through the model**, so the page is free to produce
|
|
385
378
|
however large it is; the *written*
|
|
386
379
|
review — a short read plus `flag`/`confirm`/`good` notes — is the part that costs
|
|
@@ -559,7 +552,7 @@ its verdicts and its fields.
|
|
|
559
552
|
| **Branch** | `spec/feat-orders` · 3 commits, clean |
|
|
560
553
|
| **Built** | POST /orders handler, orders schema |
|
|
561
554
|
| **Tests** | 128 passed · dotnet test |
|
|
562
|
-
| **Review** | 7 files, +212 −18 · **local** [http://127.0.0.1:7760/…](http://127.0.0.1:7760/…) · **network** [http://192.168.0.136:7760/…](http://192.168.0.136:7760/…) · **remote**
|
|
555
|
+
| **Review** | 7 files, +212 −18 · **local** [http://127.0.0.1:7760/…](http://127.0.0.1:7760/…) · **network** [http://192.168.0.136:7760/…](http://192.168.0.136:7760/…) · **remote** [http://100.126.111.34:7760/…](http://100.126.111.34:7760/…) (tailnet) |
|
|
563
556
|
| **Follow-ups** | none |
|
|
564
557
|
| **Next** | `/spec-next` → phase 3 (Auth) |
|
|
565
558
|
|