ultra-builder-pro-cli 0.26.2 → 5.0.0
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/.ultra-template/contexts/TEMPLATE.md +70 -29
- package/.ultra-template/north-star.md +35 -18
- package/.ultra-template/tasks.json +1 -0
- package/.ultra-template/test-report.json +17 -1
- package/CHANGELOG.md +235 -1
- package/README.md +210 -50
- package/README.zh-CN.md +440 -0
- package/adapters/_shared/host-profile.cjs +3 -95
- package/adapters/_shared/plugin-core.cjs +61 -5
- package/adapters/_shared/runtime-assets.cjs +66 -5
- package/adapters/zcode.js +375 -0
- package/bin/delegate-worker.cjs +2 -325
- package/bin/delegate.cjs +3 -461
- package/bin/install.js +11 -6
- package/docs/ARCHITECTURE.md +133 -31
- package/docs/ARTIFACT-AUTHORITY.md +185 -24
- package/docs/DECISIONS.md +153 -18
- package/docs/PHILOSOPHY.md +58 -7
- package/docs/PLUGIN-ISOLATION-CONTRACT.md +34 -12
- package/docs/RUNTIME-COMPAT-MATRIX.md +195 -23
- package/docs/SKILL-AUTHORING.md +19 -6
- package/docs/WORKFLOW-LIFECYCLE.md +196 -44
- package/docs/evals/adversarial-review-2026-08-14.md +84 -0
- package/docs/evals/ultra-5-0-journey-results.md +62 -0
- package/docs/evals/ultra-5-0-live-acp-conformance.md +93 -0
- package/docs/evals/ultra-5-0-route-journeys.md +145 -0
- package/docs/evals/zcode-automation-2026-08-14.md +103 -0
- package/hooks/README.md +27 -4
- package/hooks/_common.py +1678 -61
- package/hooks/adapters/zcode.py +227 -0
- package/hooks/compact_context.py +176 -17
- package/hooks/mid_workflow_recall.py +27 -5
- package/hooks/post_edit_guard.py +133 -23
- package/hooks/session_context.py +707 -22
- package/package.json +7 -3
- package/skills/ultra-change/SKILL.md +69 -12
- package/skills/ultra-change/references/change-contract.md +93 -27
- package/skills/ultra-change/references/execution-grant.md +109 -0
- package/skills/ultra-change/references/primary-transfer.md +273 -0
- package/skills/ultra-change/scripts/validate_primary_transfer.cjs +1928 -0
- package/skills/ultra-delegate/SKILL.md +103 -8
- package/skills/ultra-delegate/patches/zcode-acp-0.1.0-zcode-0.16.5.patch +184 -0
- package/skills/ultra-delegate/references/delegation-contract.md +80 -1
- package/skills/ultra-delegate/scripts/acp-conformance.cjs +287 -0
- package/skills/ultra-delegate/scripts/acp-driver.cjs +762 -0
- package/skills/ultra-delegate/scripts/agent-execution-driver.cjs +320 -0
- package/skills/ultra-delegate/scripts/cli-driver.cjs +61 -0
- package/skills/ultra-delegate/scripts/delegate-worker.cjs +328 -0
- package/skills/ultra-delegate/scripts/delegate.cjs +690 -0
- package/skills/ultra-delegate/scripts/host-profile.cjs +267 -0
- package/skills/ultra-deliver/SKILL.md +98 -11
- package/skills/ultra-dev/SKILL.md +188 -42
- package/skills/ultra-grilling/SKILL.md +10 -7
- package/skills/ultra-grilling/references/reframing.md +7 -3
- package/skills/ultra-init/SKILL.md +56 -7
- package/skills/ultra-init/scripts/init_project.cjs +968 -21
- package/skills/ultra-plan/SKILL.md +91 -47
- package/skills/ultra-plan/references/task-evidence-v2.md +300 -0
- package/skills/ultra-plan/scripts/validate_task_evidence.cjs +859 -0
- package/skills/ultra-research/SKILL.md +70 -14
- package/skills/ultra-research/references/04-product-strategy.md +11 -2
- package/skills/ultra-research/references/21-features-scope.md +8 -0
- package/skills/ultra-research/references/22-success-metrics.md +4 -3
- package/skills/ultra-research/references/99-synthesis.md +19 -2
- package/skills/ultra-research/references/north-star-v2.md +168 -0
- package/skills/ultra-research/references/wayfinding.md +8 -0
- package/skills/ultra-research/scripts/validate_north_star.cjs +1174 -0
- package/skills/ultra-review/SKILL.md +201 -27
- package/skills/ultra-review/references/adversarial-evaluation.md +31 -0
- package/skills/ultra-review/references/spec.md +5 -0
- package/skills/ultra-review/references/unified-schema.md +146 -11
- package/skills/ultra-review/references/worker-packet.md +151 -8
- package/skills/ultra-review/scripts/review_wait.py +2653 -63
- package/skills/ultra-status/SKILL.md +129 -15
- package/skills/ultra-status/references/route-contract.md +100 -0
- package/skills/ultra-tdd/SKILL.md +11 -4
- package/skills/ultra-tdd/references/test-execution.md +6 -1
- package/skills/ultra-test/SKILL.md +161 -18
- package/skills/ultra-test/scripts/validate_review_transport.cjs +549 -0
- package/skills/ultra-test/scripts/worktree_digest.cjs +518 -30
- /package/{adapters/_shared → skills/ultra-delegate/scripts}/delegate-profiles/kimi-read-only.md +0 -0
- /package/{adapters/_shared → skills/ultra-delegate/scripts}/delegate-profiles/kimi-write.md +0 -0
|
@@ -1,10 +1,5 @@
|
|
|
1
1
|
# Task {id}: {title}
|
|
2
2
|
|
|
3
|
-
> **Status**: pending | **Priority**: {priority} | **Complexity**: {complexity}
|
|
4
|
-
|
|
5
|
-
<!-- Status here and in tasks.json must always match. Both are written together, then both
|
|
6
|
-
are read back to confirm. A single-sided write breaks cross-host resume. -->
|
|
7
|
-
|
|
8
3
|
## Context
|
|
9
4
|
|
|
10
5
|
**What**: [what to build or change]
|
|
@@ -17,30 +12,54 @@
|
|
|
17
12
|
|
|
18
13
|
## Implementation
|
|
19
14
|
|
|
20
|
-
**
|
|
21
|
-
- `path/to/file.ts` (create)
|
|
22
|
-
- `path/to/existing.ts` (modify: description)
|
|
23
|
-
|
|
24
|
-
**Layers touched**: [entry point / use case / domain / persistence — must be ≥2]
|
|
15
|
+
**Layers touched**: [entry point / use case / domain / persistence]
|
|
25
16
|
|
|
26
17
|
**Pattern**: [existing code to follow]
|
|
27
18
|
|
|
28
19
|
**Tech Notes**: [framework or library specifics]
|
|
29
20
|
|
|
21
|
+
**Earliest observable slice**: [the first real consumer -> real data path ->
|
|
22
|
+
owner-visible observation, or `not applicable` with the concrete reason]
|
|
23
|
+
|
|
24
|
+
**Fresh-session recovery**: [repository root and HEAD -> active Change -> matching
|
|
25
|
+
ledger row -> this context and Resume Note -> exact safe resume command or path]
|
|
26
|
+
|
|
27
|
+
## Planned Path Inventory
|
|
28
|
+
|
|
29
|
+
`MODIFY`:
|
|
30
|
+
|
|
31
|
+
- `path/to/existing.ts`
|
|
32
|
+
|
|
33
|
+
`CREATE`:
|
|
34
|
+
|
|
35
|
+
- `path/to/file.ts`
|
|
36
|
+
|
|
37
|
+
Any later addition, removal, or rename in this inventory or the Public Seams below is
|
|
38
|
+
plan-critical and requires a rebuilt Execution Packet plus owner reapproval before
|
|
39
|
+
implementation continues.
|
|
40
|
+
|
|
41
|
+
## Public Seams
|
|
42
|
+
|
|
43
|
+
- `[entry point, schema, file contract, or other externally observable seam]`
|
|
44
|
+
|
|
45
|
+
## Narrow Verification
|
|
46
|
+
|
|
47
|
+
- `exact command for the smallest relevant contract suite`
|
|
48
|
+
- `[inspection, owner record, or external observation required by the typed criteria]`
|
|
49
|
+
|
|
30
50
|
## Acceptance Criteria
|
|
31
51
|
|
|
32
|
-
<!-- The
|
|
33
|
-
|
|
52
|
+
<!-- The typed definition of done. Readers inject this section into agent context; keep
|
|
53
|
+
every criterion observable and keep noise out. The ledger is the sole Status source. -->
|
|
34
54
|
|
|
35
55
|
**Change Acceptance IDs**: [`AC-1`, `AC-3`, or `none` for task-local-only work]
|
|
36
56
|
|
|
37
|
-
|
|
38
|
-
|
|
39
|
-
- [ ]
|
|
40
|
-
- [ ]
|
|
41
|
-
|
|
42
|
-
|
|
43
|
-
curl -X POST localhost:3000/api/example -->
|
|
57
|
+
| ID | Criterion | Verification type | Required evidence |
|
|
58
|
+
|---|---|---|---|
|
|
59
|
+
| A-01 | [observable criterion] | `command` | exact command, cwd, exit code, raw evidence ref, and freshness identity |
|
|
60
|
+
| A-02 | [observable criterion] | `inspection` | source/path, observed fact, and revision |
|
|
61
|
+
| A-03 | [owner decision when genuinely required] | `owner-judgment` | durable owner statement or explicit disposition |
|
|
62
|
+
| A-04 | [external fact when genuinely required] | `external-observation` | provider, run id, timestamp, raw evidence ref, and observation |
|
|
44
63
|
|
|
45
64
|
## Definition of Drift
|
|
46
65
|
|
|
@@ -53,10 +72,17 @@
|
|
|
53
72
|
|
|
54
73
|
**Source**: `.ultra/specs/product.md#section-id`
|
|
55
74
|
|
|
56
|
-
**
|
|
57
|
-
|
|
58
|
-
|
|
59
|
-
|
|
75
|
+
**First principles**: [`FP-1`, or every accepted principle this task relies on]
|
|
76
|
+
|
|
77
|
+
**Serves**: [`NS-1`, or every accepted outcome this task causally advances]
|
|
78
|
+
|
|
79
|
+
**Causal contribution**: [one task-local sentence connecting the planned seam to the
|
|
80
|
+
listed outcome through the listed principles; reference IDs and do not copy North Star
|
|
81
|
+
prose]
|
|
82
|
+
|
|
83
|
+
**Hard constraints**: [`HC-1`, `HC-3`, or `none` — the accepted constraints this task
|
|
84
|
+
could plausibly violate. Read them before the first edit and again before marking
|
|
85
|
+
complete. Crossing one stops the work and goes to the owner, however good the argument.]
|
|
60
86
|
|
|
61
87
|
**Decisions**: [relevant `.ultra/decisions/<id>.md` entries, or `none`]
|
|
62
88
|
|
|
@@ -77,18 +103,33 @@ the argument.]
|
|
|
77
103
|
|
|
78
104
|
## Resume Note
|
|
79
105
|
|
|
80
|
-
<!--
|
|
81
|
-
|
|
106
|
+
<!-- Navigational state only: the current checkpoint, the next action a higher
|
|
107
|
+
authority has already allowed, unresolved prerequisites, and the cheapest
|
|
108
|
+
safe resume command or path. It cannot override current owner authority,
|
|
109
|
+
approved scope/budget, task acceptance, or a validated Review verdict, it
|
|
110
|
+
cannot promote P2/P3 findings into blockers, and it never requires a fresh
|
|
111
|
+
zero-finding review after a current APPROVE. It never carries or re-issues
|
|
112
|
+
an execution grant: continuing under a durable work-package grant requires
|
|
113
|
+
the consuming Agent to re-verify the recorded grant itself. Overwrite it
|
|
114
|
+
whenever work pauses. Say where to pick up, not what was done. When the
|
|
115
|
+
pause is owner-facing, the note answers the checkpoint semantics: why,
|
|
116
|
+
outcome, accepted boundary, delta, reality, decision needed, next bounded
|
|
117
|
+
action, and not-done. -->
|
|
82
118
|
|
|
83
119
|
_(not started)_
|
|
84
120
|
|
|
85
121
|
## Completion
|
|
86
122
|
|
|
87
|
-
|
|
123
|
+
_Not completed._
|
|
124
|
+
|
|
125
|
+
## Task Review
|
|
88
126
|
|
|
89
|
-
-
|
|
90
|
-
-
|
|
91
|
-
-
|
|
127
|
+
- Execution Packet state/digest/limitation: `not_run`
|
|
128
|
+
- Review session identity and summary digest: `not_run`
|
|
129
|
+
- Blocking findings with resolution/disposition/evidence refresh refs: `not_run`
|
|
130
|
+
- Retention: retain the current strict review session until both aggregate Test and
|
|
131
|
+
Deliver consume it. Premature loss requires a fresh Review and Test; never reconstruct
|
|
132
|
+
the old receipt.
|
|
92
133
|
|
|
93
134
|
---
|
|
94
135
|
*Generated by ultra-plan*
|
|
@@ -1,35 +1,52 @@
|
|
|
1
1
|
# Project North Star
|
|
2
2
|
|
|
3
|
-
> **Authority**: the
|
|
4
|
-
> owner
|
|
3
|
+
> **Authority**: `ultra-research` is the first semantic writer. This placeholder is
|
|
4
|
+
> owner-readable structure only; it is not an accepted direction, metric, or boundary.
|
|
5
5
|
>
|
|
6
|
-
> **
|
|
7
|
-
>
|
|
8
|
-
> supporting measurements, actors, and scenarios stay in `specs/product.md`.
|
|
6
|
+
> **Adoption**: Init creates this file with status `unresearched` and no semantic IDs.
|
|
7
|
+
> Research replaces it only after evidence collection and explicit owner acceptance.
|
|
9
8
|
|
|
10
|
-
##
|
|
9
|
+
## Acceptance and Revision
|
|
11
10
|
|
|
12
|
-
|
|
11
|
+
- Schema: `north-star-v2`
|
|
12
|
+
- Status: `unresearched`
|
|
13
|
+
- Revision: `none`
|
|
14
|
+
- Owner acceptance source: `none`
|
|
15
|
+
- Acceptance time: `not-recorded`
|
|
16
|
+
- Supersedes: `none`
|
|
13
17
|
|
|
14
|
-
##
|
|
18
|
+
## Problem Reality
|
|
15
19
|
|
|
16
|
-
-
|
|
17
|
-
-
|
|
18
|
-
-
|
|
20
|
+
- Reality: [NEEDS RESEARCH]
|
|
21
|
+
- Evidence: [NEEDS RESEARCH]
|
|
22
|
+
- Unknowns: [NEEDS RESEARCH]
|
|
19
23
|
|
|
20
|
-
|
|
21
|
-
|
|
24
|
+
## First-Principle Propositions
|
|
25
|
+
|
|
26
|
+
[NEEDS RESEARCH: Research creates stable proposition IDs only after owner acceptance.]
|
|
27
|
+
|
|
28
|
+
## Value Causal Chain
|
|
29
|
+
|
|
30
|
+
[NEEDS RESEARCH: map accepted principles through capability and behavior to outcomes.]
|
|
31
|
+
|
|
32
|
+
## North Star Outcomes
|
|
33
|
+
|
|
34
|
+
[NEEDS RESEARCH: do not create an outcome ID or metric during Init.]
|
|
22
35
|
|
|
23
36
|
## Hard Constraints
|
|
24
37
|
|
|
25
|
-
|
|
38
|
+
[NEEDS RESEARCH: do not create a constraint ID during Init.]
|
|
26
39
|
|
|
27
40
|
## Explicit Exclusions
|
|
28
41
|
|
|
29
|
-
- [NEEDS
|
|
42
|
+
- [NEEDS RESEARCH]
|
|
43
|
+
|
|
44
|
+
## Uncertainties and Revisit Triggers
|
|
45
|
+
|
|
46
|
+
- [NEEDS RESEARCH]
|
|
30
47
|
|
|
31
48
|
## Research Trace
|
|
32
49
|
|
|
33
|
-
- Project
|
|
34
|
-
-
|
|
35
|
-
-
|
|
50
|
+
- Project Brief: `project-brief.md`
|
|
51
|
+
- Research runs: none
|
|
52
|
+
- Sources and decisions: none
|
|
@@ -1,7 +1,8 @@
|
|
|
1
1
|
{
|
|
2
|
-
"$schema": "ultra-test-report-
|
|
2
|
+
"$schema": "ultra-test-report-v2",
|
|
3
3
|
"change_id": null,
|
|
4
4
|
"task_ids": [],
|
|
5
|
+
"task_evidence": [],
|
|
5
6
|
"git_commit": null,
|
|
6
7
|
"intent_digest": null,
|
|
7
8
|
"worktree": {
|
|
@@ -17,6 +18,21 @@
|
|
|
17
18
|
"performance": {"status": "not_run", "evidence_refs": [], "omissions": ["Not yet run"]},
|
|
18
19
|
"security": {"status": "not_run", "evidence_refs": [], "omissions": ["Not yet run"]}
|
|
19
20
|
},
|
|
21
|
+
"review": {
|
|
22
|
+
"session": null,
|
|
23
|
+
"packet_digest": null,
|
|
24
|
+
"admission_digest": null,
|
|
25
|
+
"subject_digest": null,
|
|
26
|
+
"execution_mode": null,
|
|
27
|
+
"finding_schema": "ultra-review-findings-v4",
|
|
28
|
+
"summary_ref": null,
|
|
29
|
+
"summary_digest": null,
|
|
30
|
+
"context_digest": null,
|
|
31
|
+
"worktree_digest": null,
|
|
32
|
+
"verdict": "INCOMPLETE",
|
|
33
|
+
"coverage_refs": [],
|
|
34
|
+
"limitations": ["Not yet run"]
|
|
35
|
+
},
|
|
20
36
|
"verified_seams": [],
|
|
21
37
|
"findings": [],
|
|
22
38
|
"omissions": ["Not yet run"],
|
package/CHANGELOG.md
CHANGED
|
@@ -7,6 +7,237 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
|
|
|
7
7
|
|
|
8
8
|
## [Unreleased]
|
|
9
9
|
|
|
10
|
+
## [5.0.0] — 2026-08-26
|
|
11
|
+
|
|
12
|
+
### Added
|
|
13
|
+
|
|
14
|
+
- Added one owner-visible Route Contract for every public workflow transition:
|
|
15
|
+
invocation, mode, current position, observable exit condition, material checkpoint,
|
|
16
|
+
and next route are derived from repository authority rather than stored as another
|
|
17
|
+
semantic state machine.
|
|
18
|
+
- Added a canonical installed Skill-local delegation runtime and one replaceable
|
|
19
|
+
`AgentExecutionDriver` boundary. The stable ACP v1 driver handles negotiation, fresh
|
|
20
|
+
absolute-cwd sessions, bounded/redacted updates, permission requests, cancellation,
|
|
21
|
+
timeouts, and terminal observation while preserving the existing immutable packet,
|
|
22
|
+
Git-diff, and exact result validator. `CliDriver` remains result-compatible fallback.
|
|
23
|
+
- Added per-Host ACP capability/conformance selection and Doctor reporting for runner
|
|
24
|
+
health, selected driver, fallback reason, and transport maturity. ACP can activate
|
|
25
|
+
only for an executable stdio v1 candidate with a passed Host-specific evidence row.
|
|
26
|
+
- Added a 3.x migration and fourteen-journey acceptance harness covering byte- and
|
|
27
|
+
mode-preserving install/update/Doctor/uninstall, fresh-session recovery, degraded
|
|
28
|
+
authority, early observable slices, delegation, cancellation, and truthful STOPs.
|
|
29
|
+
|
|
30
|
+
### Changed
|
|
31
|
+
|
|
32
|
+
- Made `ultra-change` the daily entry for an existing-project request. `ultra-research`
|
|
33
|
+
now owns the first baseline or one bounded evidence question; Change owns the delta
|
|
34
|
+
and its Research Disposition; Plan owns the smallest dependency-correct graph and the
|
|
35
|
+
earliest owner-observable vertical slice.
|
|
36
|
+
- Made fresh-session recovery explicitly file-first: resolve the unique active Change,
|
|
37
|
+
task ledger and context, Acceptance, Resume Note, decisions, evidence, and Git without
|
|
38
|
+
replaying completed Research or treating conversation memory as authority.
|
|
39
|
+
- Moved the canonical delegation launcher and its complete runtime closure under the
|
|
40
|
+
installed `ultra-delegate` Skill. The package-level `ubp delegate` command is a thin
|
|
41
|
+
one-release compatibility wrapper rather than a day-to-day runtime dependency.
|
|
42
|
+
- Activated ACP-first delivery for all six shipped Host profiles only while each exact
|
|
43
|
+
executable, candidate identity, protocol version, and current conformance receipt
|
|
44
|
+
remain valid. Any unavailable or stale boundary still returns the typed,
|
|
45
|
+
result-compatible CLI fallback without an automatic provider retry.
|
|
46
|
+
- Connected ZCode Desktop through the independently built `zcode-acp` v0.1.0 bridge at
|
|
47
|
+
commit `42fe149d4b501469343c01f23ba3801832306d53` plus the shipped digest-bound ZCode
|
|
48
|
+
0.16.5 runtime-preferences patch. The verified provider path is domestic 智谱
|
|
49
|
+
BigModel Coding Plan, not international Z.AI OAuth; transport maturity remains
|
|
50
|
+
`experimental` and requires explicit acknowledgment.
|
|
51
|
+
- Kept Status and degraded Review read-only. Ambiguous authority reports the earliest
|
|
52
|
+
unsafe fact and reachable repair, retry, cancel, or abandon path without selecting an
|
|
53
|
+
intent, activating stored grant text, or writing semantic content.
|
|
54
|
+
|
|
55
|
+
### Compatibility and scope
|
|
56
|
+
|
|
57
|
+
- Preserved every seeded accepted 3.x project artifact byte and mode through the real
|
|
58
|
+
isolated install, update, Doctor, and uninstall lifecycle.
|
|
59
|
+
- Deliberately excluded a daemon, remote workflow service, semantic state machine, A2A
|
|
60
|
+
network, Team Pack, remote marketplace, shared authority store, automatic ACP-to-CLI
|
|
61
|
+
retry after a started run, and unverified Host capability promotion.
|
|
62
|
+
- This package has no service deployment target. npm publication and GitHub Release are
|
|
63
|
+
the only release surfaces; Team Pack, a daemon, remote workflow service, provider
|
|
64
|
+
purchase, credential export, force-push, and npm unpublish remain excluded.
|
|
65
|
+
|
|
66
|
+
### Verified
|
|
67
|
+
|
|
68
|
+
- Completed one bounded real ACP v1 provider session for Claude Code, Codex, OpenCode,
|
|
69
|
+
Kimi Code, Grok Build, and ZCode. Every receipt binds the exact Host candidate, clean
|
|
70
|
+
isolated Git subject, terminal worker observation, and redacted result facts.
|
|
71
|
+
- Installed 5.0.0 into all six real user Host homes after a mode-0700 recovery backup.
|
|
72
|
+
Final Doctor reports six healthy adapters, installed Skill-local ACP selection, and
|
|
73
|
+
ownership-safe ZCode uninstall/reinstall without changing the domestic-provider config.
|
|
74
|
+
- Passed the same-commit GitHub Actions matrix on Ubuntu and macOS with 661 Node tests,
|
|
75
|
+
89 Hook tests, and zero production dependency vulnerabilities, plus Windows 27/27
|
|
76
|
+
real ACP tests including cancel and timeout cleanup of an actual grandchild process.
|
|
77
|
+
- Reproduced Windows checkout EOL and worktree-alias failures before fixing them with
|
|
78
|
+
LF-bound bridge patches and filesystem device/file-ID comparison; both failed
|
|
79
|
+
predecessor runs remain retained in the release evidence.
|
|
80
|
+
|
|
81
|
+
## [3.0.2] — 2026-08-18
|
|
82
|
+
|
|
83
|
+
### Added
|
|
84
|
+
|
|
85
|
+
- Implemented the r3 North Star projection under a verified primary transfer:
|
|
86
|
+
North Star revision `north-star-v2-r3` with the new `FP-8` (Agent handover is
|
|
87
|
+
exclusive, verified, and recoverable), a new acceptance decision and immutable
|
|
88
|
+
snapshot, and the exact ZCode sole-writer durable grant
|
|
89
|
+
`ubp3-r3-zcode-2026-08-17` recorded in `.ultra/decisions/`.
|
|
90
|
+
- Added the primary-transfer contract between Agents
|
|
91
|
+
(`skills/ultra-change/references/primary-transfer.md`): an owner-granted
|
|
92
|
+
OFFER → verified ACK → sole-writer execution → frozen terminal RESULT protocol
|
|
93
|
+
over canonical files and Git, mutually exclusive with delegated workers,
|
|
94
|
+
with derived receipts under `.ultra/.runtime/handoffs/` and a mechanical
|
|
95
|
+
validator (`skills/ultra-change/scripts/validate_primary_transfer.cjs`) plus
|
|
96
|
+
authority, permission, effect, stale/revoke/interrupt/resume/cancel, and
|
|
97
|
+
missing-receipt recovery regressions (`tests/primary-transfer.test.cjs`).
|
|
98
|
+
`ultra-change`, `ultra-status`, and `ultra-delegate` consume the contract.
|
|
99
|
+
- Added the versioned closeout-transition contract
|
|
100
|
+
(`ultra-primary-transfer-closeout-v1` in
|
|
101
|
+
`skills/ultra-change/references/primary-transfer.md`): the immutable
|
|
102
|
+
reviewed subject of a newest completed v2 RESULT is separated from exactly
|
|
103
|
+
one uncommitted prescribed post-review closeout (final evidence record,
|
|
104
|
+
task-context closeout sections, ledger `completed`) through a CLOSEOUT
|
|
105
|
+
receipt that starts no review or handoff and never commits, while
|
|
106
|
+
implementation, Acceptance, PPI, pre-review-evidence, and unrelated-ledger
|
|
107
|
+
drift stays typed-stale; owner-authorized continuations between freeze and
|
|
108
|
+
closeout are recorded, bounded, and pinned. The closed task's ledger row is
|
|
109
|
+
bound ex-status (unique, `in_progress` → `completed`, every other field
|
|
110
|
+
structure-equivalent, re-read live), and `authorized_by` binds the existing
|
|
111
|
+
`ultra-external-review-receipt-v1` semantics (read-only reviewer, exact
|
|
112
|
+
task/change identity, authority and reviewed-contract refs by stable bytes,
|
|
113
|
+
subject equal to the closeout start, `approve` with no P0/P1). Regressions
|
|
114
|
+
cover the terminal
|
|
115
|
+
green path without RESULT edit or commit, structural and citation
|
|
116
|
+
rejections, drift during/after closeout (including current-row field drift),
|
|
117
|
+
continuation bounds, and preserved
|
|
118
|
+
history (`tests/primary-transfer.test.cjs`).
|
|
119
|
+
- Repaired the contract in the owner-authorized Round 1 pass
|
|
120
|
+
(`ubp3-r3-zcode-desktop-r1`): transfer validation is now phase-correct (an
|
|
121
|
+
ACK is a pre-write boundary record, so expected receiver edits never
|
|
122
|
+
self-invalidate an active or completed transfer; a superseded or v1 terminal
|
|
123
|
+
receipt stays historical), all receipt and bound-input reads are bounded
|
|
124
|
+
ordinary-file no-follow identity-checked reads (typed rejection for symlinks,
|
|
125
|
+
FIFOs, directories, oversize, and replacement), and a v2 terminal RESULT
|
|
126
|
+
binds the recomputed final HEAD, product worktree digest, exact full product
|
|
127
|
+
path inventory, and final frozen-input digests. Review-budget precedence is
|
|
128
|
+
one rule — an exact current owner grant overrides the versioned product
|
|
129
|
+
default (one initial review plus at most two P0/P1 delta reviews); P2/P3
|
|
130
|
+
never auto-extend, and the same root surviving three failed fixes stops
|
|
131
|
+
point-patching as an architecture problem. The ZCode delegate transport now
|
|
132
|
+
requires visible `--ack-experimental` acknowledgment and stamps
|
|
133
|
+
`transport_maturity`/`transport_surface` truth on every receipt and result,
|
|
134
|
+
so an app-internal CLI run can never be presented as the documented ZCode
|
|
135
|
+
Desktop interactive surface.
|
|
136
|
+
- Closed the two Round 2 misses on the same roots: transfer validation is now
|
|
137
|
+
fail-closed (an unobservable HEAD/Git is a typed `git_unavailable` error
|
|
138
|
+
instead of a green warning) and every read goes through the worktree-digest
|
|
139
|
+
tool's one shared stable snapshot primitive with parent-chain walks, exact
|
|
140
|
+
handoff-directory replay, same-subject supersession, and one finite coherent
|
|
141
|
+
digest×manifest terminal observation; every delegated ZCode terminal result
|
|
142
|
+
(finished, failed, cancelled, interrupted) now mechanically carries
|
|
143
|
+
`experimental_ack: true` when `--ack-experimental` authorized the run.
|
|
144
|
+
- Closed the final Round 3 miss on the same root: repo-wide handoff-root
|
|
145
|
+
discovery reuses the one stable directory observation — a symlinked,
|
|
146
|
+
unreadable, malformed, drifting, or oversize handoffs root (or entry) fails
|
|
147
|
+
typed and closed instead of silently reading as zero handoffs with
|
|
148
|
+
`valid: true`, while an absent root still legitimately means no transfers.
|
|
149
|
+
- Implemented the accepted Ultra Builder Pro 3.0 design as one Mode B durable
|
|
150
|
+
work-package projection: North Star revision `north-star-v2-r2` with seven first
|
|
151
|
+
principles, five observable outcomes (`NS-01`-`NS-05`), and eight hard constraints;
|
|
152
|
+
the provider-neutral Ultra Core Protocol naming across public docs; the
|
|
153
|
+
owner-facing checkpoint contract (why, outcome, accepted boundary, delta, reality,
|
|
154
|
+
decision needed, next bounded action, not-done); the per-fact artifact-authority
|
|
155
|
+
matrix with an explicit Execution Grant row and four effect classes; and the
|
|
156
|
+
optional Graph/Loop control-plane boundary documented as not integrated.
|
|
157
|
+
- Added dual-mode execution grants: session-local by default, `durable work-package`
|
|
158
|
+
by exact owner record, with stable verification, invalidation, and portable
|
|
159
|
+
handoff semantics in `skills/ultra-change/references/execution-grant.md`;
|
|
160
|
+
Status reports a recorded durable grant as inactive data awaiting verification.
|
|
161
|
+
- Added work-package review convergence bounded by an owner-visible budget: the
|
|
162
|
+
released default is at most one initial Review plus two P0/P1 delta Reviews per
|
|
163
|
+
coherent package, with explicit terminal outcomes, no self-extended budget, and
|
|
164
|
+
mandatory stop signals when repairs expose distinct root causes. Exact defaults
|
|
165
|
+
live in the versioned product contract; exact overrides in owner grants.
|
|
166
|
+
- Added ZCode as a sixth native host with fourteen Skills, five hook registrations, a
|
|
167
|
+
managed local marketplace, inline-plugin activation, Doctor, update, uninstall, and
|
|
168
|
+
bounded source/target delegation profiles. The App-bundled headless CLI transport
|
|
169
|
+
is recorded `experimental` in the shared host profile until official documentation
|
|
170
|
+
plus a full recovery drill meet the support bar.
|
|
171
|
+
- Added canonical North Star traces from Research through Change, Plan, review, Test,
|
|
172
|
+
and delivery, plus adversarial challenges at Research checkpoints 04, 21, and 99.
|
|
173
|
+
- Added review coverage references, recorded isolated-versus-sequential execution mode,
|
|
174
|
+
mandatory Plan and aggregate Test reviews, blind cross-family probes, and a seeded
|
|
175
|
+
adversarial evaluation fixture.
|
|
176
|
+
|
|
177
|
+
### Changed
|
|
178
|
+
|
|
179
|
+
- Made review topology owner-selected per stage with a one-reviewer default: initial
|
|
180
|
+
task review selects `review-spec` plus risk/touched-seam lenses, delta reviews rerun
|
|
181
|
+
only affected lenses, and the aggregate full-roster default applies only when
|
|
182
|
+
cross-task wiring justifies it — never a mandatory count or a quality proxy.
|
|
183
|
+
- Split `ultra-deliver` entry by invocation kind: a model-selected Deliver run under a
|
|
184
|
+
grant may reconcile, review, and report, then stops before finalization; writing
|
|
185
|
+
`delivery.md`, version/package posture, and archiving require a current explicit
|
|
186
|
+
owner invocation in every grant mode.
|
|
187
|
+
- Replaced the Change-scoped same-session Autonomy Envelope (superseded name,
|
|
188
|
+
historical) with the dual-mode
|
|
189
|
+
Execution Grant contract (`## Execution Grant` in every intent, `session-local`
|
|
190
|
+
default); renamed `AUTONOMY_CONTINUABLE_SKILLS` to `GRANT_CONTINUABLE_SKILLS` and
|
|
191
|
+
removed the unused `requiresAutonomyEnvelope` policy field.
|
|
192
|
+
- Closed the superseded `chg-v027-lifecycle-closure` Change with an exact
|
|
193
|
+
abandonment record citing the Mode B grant and successor `chg-ultra-3-0-mode-b`;
|
|
194
|
+
its pending v0.27 task rows remain inert append-only history.
|
|
195
|
+
- Made `validate_repo_path` in the review waiter reject symlink and non-directory
|
|
196
|
+
path components at validation time (Python 3.13 `resolve(strict=False)` no longer
|
|
197
|
+
raises on symlink loops).
|
|
198
|
+
- Made the task-evidence audit and review-wait fixtures derive the repository
|
|
199
|
+
North Star decision, snapshot, revision, and trace IDs dynamically instead of
|
|
200
|
+
locking revision r1.
|
|
201
|
+
- Raised the Grok delegation turn ceiling to twelve and retained launcher-side strict
|
|
202
|
+
result validation; Kimi delegation can now select a model explicitly.
|
|
203
|
+
- Made Claude delegation rely on strict launcher validation instead of a native schema
|
|
204
|
+
option rejected by current Claude Code.
|
|
205
|
+
- Embedded each digest-bound instruction and permission packet in the worker prompt so
|
|
206
|
+
OpenCode can keep external-directory access denied, and projected a Codex-compatible
|
|
207
|
+
native result schema while retaining stronger launcher validation.
|
|
208
|
+
|
|
209
|
+
### Fixed
|
|
210
|
+
|
|
211
|
+
- Replaced the release gate's load-sensitive fixed delay for strict Review
|
|
212
|
+
admission pinning with an explicit subprocess read handshake, so a slow clean
|
|
213
|
+
CI checkout cannot capture the already-mutated receipt and falsely pass the
|
|
214
|
+
polling invariant.
|
|
215
|
+
- Made the release gate reproducible from a clean checkout: installer integration
|
|
216
|
+
tests use an isolated Codex CLI fixture, and completed-task audits consume canonical
|
|
217
|
+
evidence after Review and Handoff receipts have reached their documented GC boundary.
|
|
218
|
+
- Made no-op ZCode hook adapters emit an actually empty stdout response, matching the
|
|
219
|
+
host's strict hook parser instead of intermittently failing a benign tool call on `{}`.
|
|
220
|
+
- Shared the macOS App-bundled ZCode CLI fallback between native plugin lifecycle
|
|
221
|
+
checks and delegated target launches when `zcode` is not available on `PATH`.
|
|
222
|
+
- Removed a delegated-result prompt contradiction by listing `$schema` in the exact
|
|
223
|
+
required field set, matching the launcher's fail-closed result validator.
|
|
224
|
+
- Added a resident entry guard to every implicitly discoverable public workflow so a
|
|
225
|
+
missing live execution grant stops before workflow work begins: a session-local grant
|
|
226
|
+
requires current conversation activation, and a durable work-package grant requires
|
|
227
|
+
stable verification by the consuming Agent.
|
|
228
|
+
- Raised the direct `js-yaml` floor to 4.3.1 so the release dependency audit no longer
|
|
229
|
+
includes the high-severity `!!omap` quadratic-consumption advisory.
|
|
230
|
+
|
|
231
|
+
### Verified
|
|
232
|
+
|
|
233
|
+
- Completed authenticated read-only target delegation with Claude Code, Codex,
|
|
234
|
+
OpenCode, Kimi Code, and ZCode; Grok Build's malformed terminal output failed closed.
|
|
235
|
+
- Re-ran ZCode target delegation without a binary override on a machine where `zcode`
|
|
236
|
+
is absent from `PATH`; the App-bundled fallback returned a validated empty-diff result.
|
|
237
|
+
- Completed one bounded ZCode automatic-coding run through Plan, adversarial reviews,
|
|
238
|
+
TDD, and Test with 3/3 tests passing, then stopped before ungranted delivery, commit,
|
|
239
|
+
push, publication, or deployment.
|
|
240
|
+
|
|
10
241
|
## [0.26.2] — 2026-08-04
|
|
11
242
|
|
|
12
243
|
### Added
|
|
@@ -1286,7 +1517,10 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
|
|
|
1286
1517
|
skill manifest, CLI protocol + mapping table; 5 spec validators.
|
|
1287
1518
|
- **Phase 0 — skeleton**: multi-runtime installer scaffolding.
|
|
1288
1519
|
|
|
1289
|
-
[Unreleased]: https://github.com/rocky2431/ultra-builder-pro-cli/compare/
|
|
1520
|
+
[Unreleased]: https://github.com/rocky2431/ultra-builder-pro-cli/compare/v5.0.0...HEAD
|
|
1521
|
+
[5.0.0]: https://github.com/rocky2431/ultra-builder-pro-cli/compare/v3.0.2...v5.0.0
|
|
1522
|
+
[3.0.2]: https://github.com/rocky2431/ultra-builder-pro-cli/compare/v0.26.2...v3.0.2
|
|
1523
|
+
[0.26.2]: https://github.com/rocky2431/ultra-builder-pro-cli/compare/v0.26.1...v0.26.2
|
|
1290
1524
|
[0.26.1]: https://github.com/rocky2431/ultra-builder-pro-cli/compare/v0.26.0...v0.26.1
|
|
1291
1525
|
[0.26.0]: https://github.com/rocky2431/ultra-builder-pro-cli/compare/v0.25.1...v0.26.0
|
|
1292
1526
|
[0.25.1]: https://github.com/rocky2431/ultra-builder-pro-cli/compare/v0.25.0...v0.25.1
|