@caddis/cli 0.4.79 → 0.4.84

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.
Files changed (60) hide show
  1. package/bundles/antigravity-plugin/plugin.json +1 -1
  2. package/bundles/antigravity-plugin/scripts/__pycache__/caddis_exit.cpython-311.pyc +0 -0
  3. package/bundles/antigravity-plugin/scripts/__pycache__/caddis_file_lock.cpython-311.pyc +0 -0
  4. package/bundles/antigravity-plugin/scripts/__pycache__/caddis_frontmatter.cpython-311.pyc +0 -0
  5. package/bundles/antigravity-plugin/scripts/__pycache__/caddis_redact.cpython-311.pyc +0 -0
  6. package/bundles/antigravity-plugin/scripts/__pycache__/caddis_roadmap.cpython-311.pyc +0 -0
  7. package/bundles/antigravity-plugin/scripts/__pycache__/caddis_roadmap_page.cpython-311.pyc +0 -0
  8. package/bundles/antigravity-plugin/scripts/__pycache__/caddis_rows.cpython-311.pyc +0 -0
  9. package/bundles/antigravity-plugin/scripts/__pycache__/caddis_tidy.cpython-311.pyc +0 -0
  10. package/bundles/antigravity-plugin/scripts/__pycache__/caddis_todo.cpython-311.pyc +0 -0
  11. package/bundles/antigravity-plugin/scripts/__pycache__/caddis_workstreams.cpython-311.pyc +0 -0
  12. package/bundles/antigravity-plugin/scripts/__pycache__/caddis_worktree_safety.cpython-311.pyc +0 -0
  13. package/bundles/antigravity-plugin/scripts/caddis_coordinate.py +128 -7
  14. package/bundles/antigravity-plugin/scripts/caddis_gate.py +3 -3
  15. package/bundles/antigravity-plugin/scripts/caddis_hub_collect.py +8 -0
  16. package/bundles/antigravity-plugin/scripts/caddis_roadmap.py +29 -3
  17. package/bundles/antigravity-plugin/scripts/caddis_roadmap_page.py +4 -3
  18. package/bundles/antigravity-plugin/scripts/caddis_workstreams.py +24 -0
  19. package/bundles/antigravity-plugin/scripts/claudster_doctor.py +24 -0
  20. package/bundles/antigravity-plugin/scripts/schemas/coordinator.schema.json +28 -1
  21. package/bundles/antigravity-plugin/scripts/schemas/hub-collect.schema.json +127 -0
  22. package/bundles/antigravity-plugin/scripts/schemas/hub-relays.schema.json +36 -0
  23. package/bundles/antigravity-plugin/scripts/schemas/hub-roadmaps.schema.json +91 -0
  24. package/bundles/antigravity-plugin/skills/coordinate/SKILL.md +17 -1
  25. package/bundles/antigravity-plugin/skills/cross-review/SKILL.md +1 -1
  26. package/bundles/antigravity-plugin/skills/feature-plan/SKILL.md +18 -9
  27. package/bundles/antigravity-plugin/skills/handoff/SKILL.md +5 -0
  28. package/bundles/antigravity-plugin/skills/implement/SKILL.md +23 -8
  29. package/bundles/antigravity-plugin/skills/mmr/SKILL.md +1 -1
  30. package/bundles/antigravity-plugin/skills/pr-merge/SKILL.md +2 -2
  31. package/bundles/antigravity-plugin/skills/ship/SKILL.md +1 -1
  32. package/bundles/antigravity-plugin/skills/ship-merge/SKILL.md +1 -1
  33. package/bundles/antigravity-plugin/skills/ship-pr/SKILL.md +2 -2
  34. package/bundles/antigravity-plugin-extras/plugin.json +1 -1
  35. package/bundles/antigravity-plugin-extras/skills/cross-review/SKILL.md +1 -1
  36. package/bundles/manifest.json +3 -3
  37. package/dist/allowed-urls.txt +19 -0
  38. package/dist/cli.js +1811 -784
  39. package/dist/export/assets/export-C0_skoDL.css +1 -0
  40. package/dist/export/assets/export-DErZHj4B.js +150 -0
  41. package/dist/export/assets/mark-small-CwCIyjyC.svg +8 -0
  42. package/dist/export/assets/wordmark-CD0s2ofS.svg +99 -0
  43. package/dist/export/fonts/LICENSE-Lucide.txt +43 -0
  44. package/dist/export/fonts/OFL-Atkinson.txt +93 -0
  45. package/dist/export/fonts/OFL-MapleMono.txt +93 -0
  46. package/dist/export/fonts/atkinson-400.849ae6f0.woff2 +0 -0
  47. package/dist/export/fonts/atkinson-700.3a9cc18a.woff2 +0 -0
  48. package/dist/export/fonts/maple-mono-400.071799fd.woff2 +0 -0
  49. package/dist/export/fonts/maple-mono-700.1e17b237.woff2 +0 -0
  50. package/dist/export/index.html +16 -0
  51. package/dist/ui/assets/Activity-CaqQ_HdD.js +6 -0
  52. package/dist/ui/assets/{Guide-D6PTiZmH.js → Guide-9PV1m0e9.js} +1 -1
  53. package/dist/ui/assets/RoadmapTab-CDm42wB2.js +31 -0
  54. package/dist/ui/assets/{index-DWiypj6N.css → index-Cv4RVfyi.css} +1 -1
  55. package/dist/ui/assets/index-Dotj3WfG.js +221 -0
  56. package/dist/ui/assets/time-D00tu2ZP.js +1 -0
  57. package/dist/ui/fonts/LICENSE-Lucide.txt +43 -0
  58. package/dist/ui/index.html +2 -2
  59. package/package.json +2 -2
  60. package/dist/ui/assets/index-D7fsc0PF.js +0 -221
@@ -71,6 +71,42 @@
71
71
  "string",
72
72
  "null"
73
73
  ]
74
+ },
75
+ "recommendation": {
76
+ "description": "the asking session's recommendation (D16), or null",
77
+ "anyOf": [
78
+ {
79
+ "type": "null"
80
+ },
81
+ {
82
+ "type": "object",
83
+ "required": [
84
+ "decision",
85
+ "reason",
86
+ "evidence"
87
+ ],
88
+ "properties": {
89
+ "decision": {
90
+ "enum": [
91
+ "approve",
92
+ "hold"
93
+ ]
94
+ },
95
+ "reason": {
96
+ "type": [
97
+ "string",
98
+ "null"
99
+ ]
100
+ },
101
+ "evidence": {
102
+ "type": [
103
+ "string",
104
+ "null"
105
+ ]
106
+ }
107
+ }
108
+ }
109
+ ]
74
110
  }
75
111
  }
76
112
  }
@@ -126,6 +126,36 @@
126
126
  },
127
127
  "state": {
128
128
  "type": "string"
129
+ },
130
+ "plan": {
131
+ "type": "string"
132
+ },
133
+ "name": {
134
+ "type": "string"
135
+ },
136
+ "model": {
137
+ "type": "string"
138
+ },
139
+ "lane": {
140
+ "type": "string"
141
+ },
142
+ "risk": {
143
+ "type": "string"
144
+ },
145
+ "tracker": {
146
+ "type": "string"
147
+ },
148
+ "tile": {
149
+ "type": "string"
150
+ },
151
+ "num": {
152
+ "type": "integer"
153
+ },
154
+ "prs": {
155
+ "type": "array",
156
+ "items": {
157
+ "type": "integer"
158
+ }
129
159
  }
130
160
  }
131
161
  }
@@ -160,6 +190,67 @@
160
190
  },
161
191
  "line": {
162
192
  "type": "string"
193
+ },
194
+ "origin_commit": {
195
+ "type": "string"
196
+ },
197
+ "lanes": {
198
+ "type": "array",
199
+ "items": {
200
+ "type": "string"
201
+ }
202
+ },
203
+ "waiting": {
204
+ "type": "array",
205
+ "items": {
206
+ "type": "object",
207
+ "properties": {
208
+ "milestone": {
209
+ "type": "string"
210
+ },
211
+ "reason": {
212
+ "type": "string"
213
+ },
214
+ "check": {
215
+ "type": [
216
+ "string",
217
+ "null"
218
+ ]
219
+ }
220
+ }
221
+ }
222
+ },
223
+ "runner": {
224
+ "type": "object",
225
+ "properties": {
226
+ "lock": {
227
+ "type": [
228
+ "object",
229
+ "null"
230
+ ]
231
+ },
232
+ "stale_lock": {
233
+ "type": [
234
+ "object",
235
+ "null"
236
+ ]
237
+ },
238
+ "last_stop": {
239
+ "type": [
240
+ "object",
241
+ "null"
242
+ ]
243
+ },
244
+ "errors": {
245
+ "type": "array",
246
+ "items": {
247
+ "type": "string"
248
+ }
249
+ }
250
+ }
251
+ },
252
+ "detail_error": {
253
+ "type": "string"
163
254
  }
164
255
  }
165
256
  },
@@ -26,6 +26,15 @@ Argument: **$ARGUMENTS**
26
26
  - After a crash: run `/caddis:coordinate start` only (first `/rename coordinator` if the name is lost).
27
27
  - Do not paste a resume prompt. Start reads the coordinator's relay and tasks itself (step 1.8).
28
28
 
29
+ ## Keep this session cheap
30
+
31
+ Measured 24-25 Oct 2026: each background wake-up re-reads the whole conversation.
32
+ - Start the coordinator on Sonnet with low effort: `claude --model sonnet --effort low --name coordinator`,
33
+ then `/caddis:coordinate start`.
34
+ - Post a peer message only when the queue or a PR state changes. Send no acknowledgement of an
35
+ acknowledgement. The watcher already prints a line only on a change, so a quiet cycle wakes nobody.
36
+ - Restart (handoff, `/clear`, `/caddis:coordinate`) when this conversation passes about 200k tokens, or once a day. The context_size hook prints a [caddis] warning at 200k and a restart notice at 350k.
37
+
29
38
  ## How to call the script
30
39
 
31
40
  - Run these commands with the Bash tool: the quoting and the `GH_TOKEN=$(...)` forms are POSIX shell.
@@ -209,7 +218,7 @@ and 7.3. While the queue is frozen (section 5 step 1), do steps 3.1 to 3.3 only.
209
218
  GH_TOKEN=$(gh auth token -u <account>) python "${CLAUDE_PLUGIN_ROOT}/scripts/caddis_lanes.py" ci-watch --pr <n>
210
219
  ```
211
220
  Act only on `RESULT: pass`. The `HEAD:` line is the `<watched sha>`. Exit 1: CI is red (design 8.1;
212
- one re-run for a flake, design 8.2). Exit 3: run `ci-watch` again; if it stays 3, stop and report.
221
+ one re-run for a flake, design 8.2). Exit 3: watch the queued or unfinished run once with `ci-watch --run <id> --repo <owner/name>` (not a new `--pr` watch); if it stays 3, stop and report. One `ci-watch` per PR head; write no "still waiting" turns.
213
222
  Exit 4: a usage error; stop and report.
214
223
  7. If the item is in `ci` (after step 3.5 or a new head), check R1 and R2 again at the watched head,
215
224
  then run `queue ready` (step 3.3). Do not warn for an item that is not `ready`.
@@ -334,6 +343,13 @@ Stalls (design 7.4): one nudge for a busy, silent session that holds a queue ite
334
343
  reason: <one or two sentences>
335
344
  evidence: <cost estimate | test or CI result | risk | how to undo>
336
345
  ```
346
+ Also record it, so the hub Inbox shows it with Approve and Decline (add `--pr <n> --head-sha <full sha>` when it
347
+ is about one PR; that PR must be in the queue):
348
+ ```
349
+ python "${CLAUDE_PLUGIN_ROOT}/scripts/caddis_coordinate.py" ask add --repo <R> --what '<one line>' --recommendation approve --reason '<text>' --evidence '<text>' --json
350
+ ```
351
+ A hub click reaches intake and removes the ask. When the owner answers in this window instead, record the
352
+ answer and run `ask remove --repo <R> --id <ask id>`.
337
353
  3. A recommendation is never an approval. Only the owner's click in the hub or the owner's own words in
338
354
  this window approve. Record the owner's answer with `ledger record --by owner --window <N>`.
339
355
  4. Hotfix marks come only from the owner (design 2). You may recommend one.
@@ -111,7 +111,7 @@ a reason.
111
111
  2. **Run it before review.** The coder runs only the tests of the files it touched, in the foreground, each command under 10 minutes (split into parts if needed). Name the exact commands in the task. Never start the full suite in the background and never wait on a monitor for it: CI runs the full suite on every PR. Then run the change once on real data.
112
112
  3. **Review by tier.** Run `python "${CLAUDE_PLUGIN_ROOT}/scripts/caddis_gate.py" review-tier --range <base>...HEAD` (add `--plan <plan> --phase <N>` inside a plan; no `TIER:` line means tier 2).
113
113
  Tier 0: gates only. Tier 1-small: your self-check only. Tier 1: one review by another vendor; stop at
114
- the first CLEAN. Tier 2: two reviewers of two other vendors, in one round. Fix blocking findings only;
114
+ the first CLEAN. Tier 2: one review by another vendor (two only when the owner asks). Fix blocking findings only;
115
115
  park the rest. Re-review only the fix diff, and only when the fix changed logic. After 3 rounds, ask the owner.
116
116
 
117
117
  A diff over the batch ceiling is split into batches. Each batch prompt lists every file in the
@@ -147,8 +147,8 @@ froze CI on the dev box.
147
147
 
148
148
  - **review lane (every phase that touches code)** — cross-vendor review by a vendor that did NOT
149
149
  write the code: `/caddis:cross-review --range main..HEAD`
150
- (Claude or GLM implemented) or `--provider glm` / a Claude `code-reviewer` pass (DeepSeek-adjacent
151
- work). Never same-vendor.
150
+ (Claude or GLM implemented) or `--provider glm` (DeepSeek-adjacent work). Never same-vendor. Never a
151
+ Claude reviewer subagent, unless every other-vendor provider failed; the executor says so in the review file.
152
152
 
153
153
  **The `Lane:` line is a routing directive read by `/caddis:implement`, not a note.** An interactive
154
154
  session that is handed a phase belonging to another lane spawns that lane rather than absorbing the
@@ -196,9 +196,10 @@ Write one `Execution:` line per phase, next to its `Lane:` and model lines. A fi
196
196
  - Otherwise `Execution: inline`.
197
197
 
198
198
  `inline` means today's behaviour: one implementer does the whole phase. `per-task` means the executor
199
- gives each task a fresh implementer (on the phase's model tier and lane) and a short task review (spec
200
- compliance, then quality) before the next task starts. The task review does not replace the phase's review by another vendor:
201
- `review-tier` and `/caddis:cross-review` still run once per phase.
199
+ gives each task a fresh implementer (on the phase's model tier and lane). There is no model review
200
+ between tasks. Each phase gets ONE review, by ONE reviewer of another vendor, at the end of the phase
201
+ (`review-tier` and `/caddis:cross-review`). A fix is re-checked only when it fixes a blocking finding
202
+ and changes logic.
202
203
 
203
204
  A `per-task` phase is split into tasks. Each task must have:
204
205
  - **`Files:`** — every file to create, modify and test, with exact paths.
@@ -278,11 +279,19 @@ to the phase or task that owns the code. Write "none found" only after you check
278
279
 
279
280
  ## Phases
280
281
 
282
+ **Session rules for the orchestrator** (measured 24-25 Oct 2026: long Opus sessions re-read 400-600k tokens
283
+ on every turn). Keep the context small: subagent replies under about 200 words, full reports in files,
284
+ never paste or tail large outputs. At each phase end, write the handoff; continue in the same session only
285
+ while the conversation is under about 200k tokens. The context_size hook prints a [caddis] warning at 200k and a restart notice at 350k. Run the phase review as
286
+ `/caddis:cross-review --range <base>..HEAD`. Send a task marked mechanical (literal tests, moves, deletes,
287
+ renames) to agy through `caddis_spend.py run`. Run one `ci-watch` per PR head; on a queued-run timeout watch
288
+ that run once with `--run <id>`; write no "still waiting" turns.
289
+
281
290
  ### Phase 1 — <name> ⏳
282
291
  > ⚠️ **Switch model/lane BEFORE starting this phase** — the *active* model does the work, not the
283
292
  > one named here. Lane `claude`: run `/model <alias>`. Lane `glm-headless`: run the exact command below.
284
293
  **Milestone:** M1 *(omit on plans without milestones)*
285
- **Model:** <tier> (`/model <alias>`) — <one-line rationale tied to this phase's difficulty>
294
+ **Model:** <tier> (`/model <alias>`) — <one-line rationale tied to this phase's difficulty>. Orchestrator: Sonnet to run the plan (Opus only to write plans and designs).
286
295
  **Lane:** claude — this session | glm-headless — `claude-glm -p "/caddis:implement <this plan's path> — Phase 1 only"`
287
296
  **Type:** backend | hard | ui | tests | docs *(required when the Lane runs through caddis_lanes.py)*
288
297
  **Execution:** inline | per-task *(the rule in Step 2; per-task phases list their tasks below)*
@@ -322,13 +331,13 @@ Three rules: one attempt, one review.
322
331
  2. **Run it before review.** The coder runs only the tests of the files it touched, in the foreground, each command under 10 minutes (split into parts if needed). Name the exact commands in the task. Never start the full suite in the background and never wait on a monitor for it: CI runs the full suite on every PR. Then run the change once on real data.
323
332
  3. **Review by tier.** Run `python "${CLAUDE_PLUGIN_ROOT}/scripts/caddis_gate.py" review-tier --range <base>...HEAD` (add `--plan <plan> --phase <N>` inside a plan; no `TIER:` line means tier 2).
324
333
  Tier 0: gates only. Tier 1-small: your self-check only. Tier 1: one review by another vendor; stop at
325
- the first CLEAN. Tier 2: two reviewers of two other vendors, in one round. Fix blocking findings only;
334
+ the first CLEAN. Tier 2: one review by another vendor (two only when the owner asks). Fix blocking findings only;
326
335
  park the rest. Re-review only the fix diff, and only when the fix changed logic. After 3 rounds, ask the owner.
327
336
 
328
337
  - One phase at a time: read this plan → run the phase on its **Lane** → RED→GREEN→VERIFY→CROSS-REVIEW→
329
338
  COMMIT → update the **Tracker** row. The Tracker is the resume signal for every future session.
330
- An `Execution: per-task` phase runs task by task: a fresh implementer per task, and a task review
331
- before the next task. The cross-review still runs once for the phase.
339
+ An `Execution: per-task` phase runs task by task: a fresh implementer per task, and no model
340
+ review between tasks. One cross-review, by one other vendor, runs at the end of the phase.
332
341
  - **Session boundaries:** the executor WRITES the handoff (durable `relay.md` + Tracker) after any
333
342
  phase that ends a sitting — it performs the work, it does not suggest the command. Start a FRESH
334
343
  session (or `/clear`) before any phase marked `Session: fresh`, before a lane change, or whenever
@@ -464,6 +464,11 @@ claude --name <role>-<MMDD>-<HH> '<the resume prompt on one line> <the announce
464
464
  `--role <role>` to set another.
465
465
 
466
466
  ## Rules
467
+ - **Clear above 200k tokens.** A session whose conversation passes about 200k tokens writes this handoff at
468
+ the next safe point and asks the owner for `/clear`. Each turn re-reads the whole conversation, so a long
469
+ session costs more with every turn. The context_size hook prints a [caddis] warning at 200k and a restart notice at 350k.
470
+ - **Model for the next session.** Print one line `Model for next session: ...` just before the RESUME block (the block stays the last output): Sonnet (effort medium) to run an approved
471
+ plan; Opus to write a plan or design. Start line: `claude --model sonnet --effort medium --name <role>-<MMDD>-<HH>`.
467
472
  - **One next action, by priority ladder.** `## Next step` names exactly ONE action — not a menu. Pick it in
468
473
  order: (1) interrupted mid-phase work → finish + commit it; else (2) the active plan's next not-started
469
474
  phase; else (3) the top open question/decision. Everything else is context, not the next step.
@@ -55,12 +55,28 @@ depend on them:
55
55
  AskUserQuestion. Where the plan leaves a genuine gap, make the smallest reasonable assumption, record it
56
56
  in the review file, and proceed — asking is always wrong here.
57
57
 
58
+ ## Keep the session small and quiet
59
+
60
+ Measured 24-25 Oct 2026: about 60% of usage was cache re-reads in long Opus main sessions (400-600k tokens),
61
+ and 95-154 turns per session were started by background events. Each turn re-reads the whole conversation.
62
+ - **Small context.** Ask each subagent for a reply under about 200 words. Full reports go to files; read a
63
+ file only for the part you need. Never paste or tail a large output.
64
+ - **Handoff at phase end.** At each phase end, write the handoff yourself (see "At every halt" below). Continue
65
+ in the same session only while the conversation is under about 200k tokens. Above that, ask for `/clear`. The context_size hook prints a [caddis] warning at 200k and a restart notice at 350k.
66
+ - **Model.** The orchestrator runs on Sonnet. Opus stays for plans, designs and critical-code subagents.
67
+ - **One phase review.** Run `/caddis:cross-review --range <base>..HEAD` (one other vendor; agy Argon by default).
68
+ Never use a Claude reviewer subagent. Exception: every other-vendor provider failed. Then say so in the review file.
69
+ - **Mechanical tasks** (literal tests, moves, deletes, renames) go to agy through
70
+ `python "${CLAUDE_PLUGIN_ROOT}/scripts/caddis_spend.py" run --caller lane-pN -- agy ...`.
71
+ - **Fewer wake-ups.** Run one `ci-watch` per PR head. When it times out on a queued run, watch that run once
72
+ with `ci-watch --run <id>`; do not start a new PR watch. Write no "still waiting" turns.
73
+
58
74
  ## Three rules: one attempt, one review
59
75
  1. **Think before you send.** Before a task goes to any coder, write its edge cases and check the facts it depends on (one query or one file read).
60
76
  2. **Run it before review.** The coder runs only the tests of the files it touched, in the foreground, each command under 10 minutes (split into parts if needed). Name the exact commands in the task. Never start the full suite in the background and never wait on a monitor for it: CI runs the full suite on every PR. Then run the change once on real data.
61
77
  3. **Review by tier.** Run `python "${CLAUDE_PLUGIN_ROOT}/scripts/caddis_gate.py" review-tier --range <base>...HEAD` (add `--plan <plan> --phase <N>` inside a plan; no `TIER:` line means tier 2).
62
78
  Tier 0: gates only. Tier 1-small: your self-check only. Tier 1: one review by another vendor; stop at
63
- the first CLEAN. Tier 2: two reviewers of two other vendors, in one round. Fix blocking findings only;
79
+ the first CLEAN. Tier 2: one review by another vendor (two only when the owner asks). Fix blocking findings only;
64
80
  park the rest. Re-review only the fix diff, and only when the fix changed logic. After 3 rounds, ask the owner.
65
81
 
66
82
  **Author self-check** (tier 1-small, and before any review). It is not a model review. Answer all five:
@@ -241,12 +257,11 @@ report — do not redo completed work.
241
257
  run the phase's tasks in order. Give each task a fresh implementer: a subagent on the phase's
242
258
  lane and model, with a brief that holds the task's `Files:`, `Interfaces:` and steps, the plan's
243
259
  Global Constraints and the test-economy lines (Multi-lane item 7b). If you cannot dispatch a
244
- subagent, implement the task yourself and note that in the review file. After each task, run a
245
- short task review before the next task: first spec compliance (the task's files, names and tests
246
- match the plan), then quality (the author self-check). Fix blocking findings, then go on. Record
247
- each task (name, commit, review result) under `## Task progress` in the review file. Then run the
248
- phase's normal gates: VERIFY, the `review-tier` review by another vendor, COMMIT and the Tracker.
249
- The task reviews never replace that review.
260
+ subagent, implement the task yourself and note that in the review file. After each task, run
261
+ the author self-check (no model review between tasks). Record each task (name, commit, self-check
262
+ result) under `## Task progress` in the review file. Then run the phase's normal gates: VERIFY,
263
+ the ONE `review-tier` review by one other vendor at the end of the phase, COMMIT and the Tracker.
264
+ Re-check a fix only when it fixes a blocking finding and changes logic.
250
265
  - **RED** — write the failing test(s) the phase names (`<test file>::<case>`). Run them; confirm they
251
266
  fail for the right reason (the missing behavior, not an import error). If a phase genuinely has no
252
267
  testable surface, say so in the review file and implement the minimal change directly.
@@ -360,7 +375,7 @@ branch: <current branch>
360
375
  - <any gap the plan left that you decided — or "none">
361
376
 
362
377
  ## Task progress *(only for `Execution: per-task` phases)*
363
- - <phase N task N.M — commit <hash> — task review: clean | fixed <finding>>
378
+ - <phase N task N.M — commit <hash> — self-check: clean | fixed <finding>>
364
379
 
365
380
  ## Not done / follow-ups
366
381
  - <phases skipped/blocked and why — or "none; all phases complete">
@@ -25,7 +25,7 @@ The exit codes follow `caddis_exit.py`: 0 clean, 1 blocked, 2 advisory, 3 could
25
25
  2. **Run it before review.** The coder runs only the tests of the files it touched, in the foreground, each command under 10 minutes (split into parts if needed). Name the exact commands in the task. Never start the full suite in the background and never wait on a monitor for it: CI runs the full suite on every PR. Then run the change once on real data.
26
26
  3. **Review by tier.** Run `python "${CLAUDE_PLUGIN_ROOT}/scripts/caddis_gate.py" review-tier --range <base>...HEAD` (add `--plan <plan> --phase <N>` inside a plan; no `TIER:` line means tier 2).
27
27
  Tier 0: gates only. Tier 1-small: your self-check only. Tier 1: one review by another vendor; stop at
28
- the first CLEAN. Tier 2: two reviewers of two other vendors, in one round. Fix blocking findings only;
28
+ the first CLEAN. Tier 2: one review by another vendor (two only when the owner asks). Fix blocking findings only;
29
29
  park the rest. Re-review only the fix diff, and only when the fix changed logic. After 3 rounds, ask the owner.
30
30
 
31
31
  `/mmr` is for handing a task to another model, not for normal reviews. For a normal review, use
@@ -95,7 +95,7 @@ hands back here.
95
95
 
96
96
  <!-- shared:cross-review-trigger — keep byte-identical across commands; a test pins it -->
97
97
  Run `python "${CLAUDE_PLUGIN_ROOT}/scripts/caddis_gate.py" review-tier --range <base>...HEAD` and review by its tier (rule 3).
98
- `<base>` is `origin/<default branch>` (on the default branch itself, the commit before this change). Tier 0 or 1-small: no model review. Tier 1: one `/caddis:cross-review`. Tier 2: two reviewers of two other vendors in one round. No `TIER:` line means tier 2.
98
+ `<base>` is `origin/<default branch>` (on the default branch itself, the commit before this change). Tier 0 or 1-small: no model review. Tier 1: one `/caddis:cross-review`. Tier 2: one `/caddis:cross-review` by one other vendor (two only when the owner asks). No `TIER:` line means tier 2.
99
99
  <!-- /shared:cross-review-trigger -->
100
100
 
101
101
  ## Step 4 — Rebase-safe currency
@@ -128,7 +128,7 @@ Watch the PR's checks job-by-job until completion:
128
128
  python "${CLAUDE_PLUGIN_ROOT}/scripts/caddis_lanes.py" ci-watch --pr <pr>
129
129
  ```
130
130
  `RESULT: pass` means every check of the PR's head commit is complete and green (cancelled, failed or unfinished is never a pass). `HEAD:` names that commit: it is the `<watched-sha>`; merge with `--match-head-commit <watched-sha>`.
131
- Exit codes: **0** pass → continue; **1** fail → read `FAILING:` and the gh output, classify the cause, apply the minimum source fix, and re-push (stop before merging); **3** could not tell (also: gh finished but a check is still unfinished, the head's checks could not be read, or the head moved during the watch — `FAILING:` says which) → run `ci-watch` again; if it stays 3, stop and report, do not merge.
131
+ Exit codes: **0** pass → continue; **1** fail → read `FAILING:` and the gh output, classify the cause, apply the minimum source fix, and re-push (stop before merging); **3** could not tell (also: gh finished but a check is still unfinished, the head's checks could not be read, or the head moved during the watch — `FAILING:` says which) → if `FAILING:` names a queued or unfinished run, watch that run once with `ci-watch --run <id>` (never a new `--pr` watch); if it stays 3, stop and report, do not merge. One `ci-watch` per PR head; write no "still waiting" turns.
132
132
  Fallback when `caddis_lanes.py` is not available: watch in-session with `gh pr checks <pr> --watch` / `gh run watch` (`gh-cli` skill).
133
133
  - If any check fails (red check): report the failure, classify the cause, and **stop before merging**.
134
134
  Do not merge a failing PR. Apply the minimum source fix and re-push.
@@ -69,7 +69,7 @@ Use conventional commits: `fix:`, `feat:`, `chore:`, `refactor:`, `docs:`. Scope
69
69
 
70
70
  <!-- shared:cross-review-trigger — keep byte-identical across commands; a test pins it -->
71
71
  Run `python "${CLAUDE_PLUGIN_ROOT}/scripts/caddis_gate.py" review-tier --range <base>...HEAD` and review by its tier (rule 3).
72
- `<base>` is `origin/<default branch>` (on the default branch itself, the commit before this change). Tier 0 or 1-small: no model review. Tier 1: one `/caddis:cross-review`. Tier 2: two reviewers of two other vendors in one round. No `TIER:` line means tier 2.
72
+ `<base>` is `origin/<default branch>` (on the default branch itself, the commit before this change). Tier 0 or 1-small: no model review. Tier 1: one `/caddis:cross-review`. Tier 2: one `/caddis:cross-review` by one other vendor (two only when the owner asks). No `TIER:` line means tier 2.
73
73
  <!-- /shared:cross-review-trigger -->
74
74
 
75
75
  **3. PUSH** — only when Step 1b found a non-default branch (on the default branch, `/ship` stopped at Step 2):
@@ -118,7 +118,7 @@ The merge lands on the default branch and triggers the pipeline. Watch it via th
118
118
  ```bash
119
119
  python "${CLAUDE_PLUGIN_ROOT}/scripts/caddis_lanes.py" ci-watch --run <run-id>
120
120
  ```
121
- Exit codes: **0** pass → continue; **1** fail → read `FAILING:` and the gh output, classify, apply the minimum source fix; **3** could not tell → stop, report, and **skip cleanup** (Step 6).
121
+ Exit codes: **0** pass → continue; **1** fail → read `FAILING:` and the gh output, classify, apply the minimum source fix; **3** could not tell → stop, report, and **skip cleanup** (Step 6). This step already watches the run with `--run`, so do not watch it again; write no "still waiting" turns.
122
122
  Fallback when `caddis_lanes.py` is not available: watch in-session with `gh run watch <run-id> --exit-status` on the default branch (`gh-cli` skill). If the workflow has a deploy job, watch it through; if CI-only, note that no deploy occurs.
123
123
 
124
124
  ## Step 5 — Post-deploy validation
@@ -124,7 +124,7 @@ source checkout; skip this step if the script is missing or the repo has no `.ca
124
124
 
125
125
  <!-- shared:cross-review-trigger — keep byte-identical across commands; a test pins it -->
126
126
  Run `python "${CLAUDE_PLUGIN_ROOT}/scripts/caddis_gate.py" review-tier --range <base>...HEAD` and review by its tier (rule 3).
127
- `<base>` is `origin/<default branch>` (on the default branch itself, the commit before this change). Tier 0 or 1-small: no model review. Tier 1: one `/caddis:cross-review`. Tier 2: two reviewers of two other vendors in one round. No `TIER:` line means tier 2.
127
+ `<base>` is `origin/<default branch>` (on the default branch itself, the commit before this change). Tier 0 or 1-small: no model review. Tier 1: one `/caddis:cross-review`. Tier 2: one `/caddis:cross-review` by one other vendor (two only when the owner asks). No `TIER:` line means tier 2.
128
128
  <!-- /shared:cross-review-trigger -->
129
129
 
130
130
  > **Why this is mechanical and not "consider a review here".** a fleet app shipped four
@@ -168,7 +168,7 @@ scope-guard check when present:
168
168
  python "${CLAUDE_PLUGIN_ROOT}/scripts/caddis_lanes.py" ci-watch --pr <pr>
169
169
  ```
170
170
  `RESULT: pass` means every check of the PR's head commit is complete and green; `HEAD:` names that commit.
171
- Exit codes: **0** pass → continue; **1** fail → read `FAILING:` and the gh output, classify, apply the minimum source fix, then re-push (through Step 4's safety if history must change); **3** could not tell (also: a check still unfinished, or the head's checks unreadable) → run `ci-watch` again; if it stays 3, stop and report, do not merge.
171
+ Exit codes: **0** pass → continue; **1** fail → read `FAILING:` and the gh output, classify, apply the minimum source fix, then re-push (through Step 4's safety if history must change); **3** could not tell (also: a check still unfinished, or the head's checks unreadable) → watch the queued or unfinished run once with `ci-watch --run <id>` (never a new `--pr` watch); if it stays 3, stop and report, do not merge. One `ci-watch` per PR head; write no "still waiting" turns.
172
172
  Fallback when `caddis_lanes.py` is not available: watch in-session with `gh pr checks <pr> --watch` / `gh run watch` (`gh-cli` skill).
173
173
  On a red check, classify the failure and apply the minimum **source** fix, then re-push (through
174
174
  Step 4's safety if history must change).
@@ -1,5 +1,5 @@
1
1
  {
2
2
  "name": "caddis-extras",
3
- "version": "1.3.46",
3
+ "version": "1.3.47",
4
4
  "description": "caddis-extras — the long-tail skills for the Antigravity (agy) caddis plugin; install alongside caddis for the full set."
5
5
  }
@@ -99,7 +99,7 @@ smaller chunks; don't just raise `--max-diff-chars` without knowing the endpoint
99
99
  2. **Run it before review.** The coder runs only the tests of the files it touched, in the foreground, each command under 10 minutes (split into parts if needed). Name the exact commands in the task. Never start the full suite in the background and never wait on a monitor for it: CI runs the full suite on every PR. Then run the change once on real data.
100
100
  3. **Review by tier.** Run `python "${CLAUDE_PLUGIN_ROOT}/scripts/caddis_gate.py" review-tier --range <base>...HEAD` (add `--plan <plan> --phase <N>` inside a plan; no `TIER:` line means tier 2).
101
101
  Tier 0: gates only. Tier 1-small: your self-check only. Tier 1: one review by another vendor; stop at
102
- the first CLEAN. Tier 2: two reviewers of two other vendors, in one round. Fix blocking findings only;
102
+ the first CLEAN. Tier 2: one review by another vendor (two only when the owner asks). Fix blocking findings only;
103
103
  park the rest. Re-review only the fix diff, and only when the fix changed logic. After 3 rounds, ask the owner.
104
104
 
105
105
  A diff over the batch ceiling is split into batches. Each batch prompt lists every file in the
@@ -1,8 +1,8 @@
1
1
  {
2
- "poolVersion": "1.3.163",
2
+ "poolVersion": "1.3.168",
3
3
  "bundles": {
4
- "antigravity-plugin": "1.3.163",
5
- "antigravity-plugin-extras": "1.3.46",
4
+ "antigravity-plugin": "1.3.168",
5
+ "antigravity-plugin-extras": "1.3.47",
6
6
  "codex": "unknown"
7
7
  }
8
8
  }
@@ -0,0 +1,19 @@
1
+ # URLs that may appear in the Pages export (R62 d). One exact URL per line.
2
+ # Any other URL in any exported file fails the privacy scan.
3
+ # Add a line only after you check that the URL is a namespace or licence link, not a private host.
4
+
5
+ # XML and SVG namespace names and the error-page link in the React bundle
6
+ http://www.w3.org/1998/Math/MathML
7
+ http://www.w3.org/1999/xhtml
8
+ http://www.w3.org/1999/xlink
9
+ http://www.w3.org/2000/svg
10
+ http://www.w3.org/XML/1998/namespace
11
+ https://reactjs.org/docs/error-decoder.html?invariant=
12
+
13
+ # Licence texts in licenses/
14
+ https://github.com/googlefonts/atkinson-hyperlegible-next
15
+ https://github.com/subframe7536/maple-font
16
+ https://openfontlicense.org
17
+
18
+ # A PR link is built from `--github` and a number at run time. The bundle holds only the template.
19
+ template: https://github.com/