ultracode-for-codex 0.2.6 → 0.3.1

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.
@@ -7,7 +7,7 @@ Date: 2026-06-22
7
7
  This audit checked:
8
8
 
9
9
  - tracked repository files;
10
- - generated npm package contents for `ultracode-for-codex@0.2.6`;
10
+ - generated npm package contents for `ultracode-for-codex@0.3.0`;
11
11
  - the locally installed companion Codex skill.
12
12
 
13
13
  Generated build output and package tarballs were checked as projections of the
@@ -23,7 +23,8 @@ License transition completed:
23
23
 
24
24
  - Apache-2.0 `LICENSE` file is present;
25
25
  - `package.json` and `package-lock.json` declare `Apache-2.0`;
26
- - package version is prepared as `0.2.6`.
26
+ - release-candidate package version is `0.3.0`;
27
+ - npm latest before this release remains `0.2.6`.
27
28
 
28
29
  ## Evidence
29
30
 
package/package.json CHANGED
@@ -1,7 +1,7 @@
1
1
  {
2
2
  "name": "ultracode-for-codex",
3
- "version": "0.2.6",
4
- "description": "Run local Codex-backed workflows from a command-owned CLI runtime.",
3
+ "version": "0.3.1",
4
+ "description": "Run Codex-native Ultracode orchestration with an optional local CLI runtime.",
5
5
  "keywords": [
6
6
  "codex",
7
7
  "workflow",
@@ -41,6 +41,7 @@
41
41
  "dist/codex/",
42
42
  "dist/runtime/",
43
43
  "skills/ultracode-for-codex/",
44
+ "skills/ultracode-for-codex-cli/",
44
45
  "settings.json",
45
46
  "README.md",
46
47
  "docs/provenance-audit.md",
@@ -58,6 +59,7 @@
58
59
  "publish:dry-run": "npm publish --dry-run --access public",
59
60
  "publish:npm": "npm publish --access public",
60
61
  "publish:npm:provenance": "npm publish --access public --provenance",
62
+ "smoke:live": "node scripts/live-smoke-ultracode-for-codex.mjs",
61
63
  "test:e2e:ultracode-for-codex": "node scripts/e2e-installed-ultracode-for-codex.mjs",
62
64
  "test:all": "npm test && npm run test:e2e:ultracode-for-codex",
63
65
  "test": "npm run build && npm run verify:runtime-boundary && node --test --test-concurrency=2 test/*.test.mjs",
@@ -1,130 +1,114 @@
1
1
  ---
2
2
  name: ultracode-for-codex
3
- description: Operate, package, validate, or update the Ultracode for Codex npm runtime. Use when installing or running local workflows, checking runtime boundaries, packaging release tarballs, updating docs, or maintaining the companion Codex skill.
3
+ description: Run Ultracode for Codex in Codex-native mode, with the main Codex context planning phases, spawning parallel subagents, synthesizing results, and showing progress directly in the chat.
4
4
  ---
5
5
 
6
6
  # Ultracode for Codex
7
7
 
8
8
  ## Core Rule
9
9
 
10
- Treat the npm package as the runtime artifact. This skill is only a companion
11
- guide for Codex agents. Runtime authority remains in the `ultracode-for-codex`
12
- binary, tests, package exports, journal layer, and workflow runtime code.
13
-
14
- Workflow execution runs through the local CLI command. Progress,
15
- cancellation, permission review, retry, and result projection stay in that
16
- command process. `settings.json` defaults runs to OS background execution; use
17
- that path for long Codex-launched work so Codex can keep doing other tasks and
18
- inspect `progressPath` or `resultPath` later. Attached runs stream stderr JSONL
19
- for Codex-readable status, while stdout remains the final workflow result JSON.
20
-
21
- The default Ultracode work shape is phase-wise parallel execution: built-in
22
- `task` and `code-review` first call a planner agent, then execute each planned
23
- phase with parallel focused subagents by default, followed by phase and final
24
- synthesis. A single-agent path is reserved for cases where the planner judges
25
- parallel execution risky or wasteful.
26
- Planner guidance includes classify-and-act, fan-out-and-synthesize,
27
- adversarial verification, generate-and-filter, tournament, and loop-until-done
28
- patterns so workflow shape can follow the task instead of a fixed template.
29
-
30
- ## Install And Run
31
-
32
- Use the npm package for consumer installs.
33
-
34
- ```bash
35
- npm install --save-dev ultracode-for-codex
36
- npm exec -- ultracode-for-codex --llm-guide
37
- npm exec -- ultracode-for-codex run \
38
- --accept-llm-guide=v1 \
39
- --cwd /path/to/project \
40
- --script-file .codex/workflows/review.js \
41
- --args '{"prompt":"review the current change"}'
42
- ```
10
+ This skill is the primary Codex-native Ultracode command. Treat the current
11
+ Codex main context as the orchestrator. Plan adaptive phases, spawn focused
12
+ subagents directly from Codex, synthesize phase outputs, and keep the user
13
+ informed in the chat.
43
14
 
44
- For source-checkout validation before publish:
15
+ Use the CLI runtime only when the user explicitly asks for `$ultracode-for-codex-cli`,
16
+ CLI execution, background jobs, packaging, publish preparation, installed
17
+ runtime validation, or reproducible local runtime artifacts.
45
18
 
46
- ```bash
47
- npm run pack:ultracode-for-codex
48
- npm install --save-dev ./artifacts/ultracode-for-codex-0.2.6.tgz
49
- ```
19
+ ## Required Capability Surface
50
20
 
51
- CLI behavior:
52
-
53
- - `--version` or `-v` prints the installed package version;
54
- - default execution is `background`; stdout contains a launch record with
55
- `jobId`, `pid`, `resultPath`, `progressPath`, `metadataPath`, and `pidPath`;
56
- - attached execution is available with `--execution attached` when the caller
57
- should stay connected until completion;
58
- - attached progress prints to stderr as JSONL by default;
59
- - attached final workflow result prints as JSON to stdout;
60
- - JSONL records include `kind`, `version`, `event`, `status`, and `summary`,
61
- with agent identity and label fields on agent records;
62
- - built-in `task` and `code-review` emit `workflow.plan.ready` as a planning
63
- snapshot, not a promise that every later phase is already known;
64
- - `workflow.phase.planned` is emitted immediately before each phase starts and
65
- carries that phase's current planned agent role labels;
66
- - each `workflow.phase.started` record repeats the same role labels when the
67
- phase begins;
68
- - each `workflow.agent.completed` record includes phase progress, total known
69
- agent progress, and elapsed time;
70
- - `Ctrl-C` cancels the active attached workflow;
71
- - `--retry-limit <n>` retries failed workflows inside the same process;
72
- - `--timeout-ms 0` waits for completion, cancellation, or app-server exit;
73
- positive values opt into a workflow deadline and per-agent silence budget,
74
- and that budget is not divided by the retry budget.
75
- - `--permission ask|allow|deny` handles project/user/plugin/scriptPath reviews.
76
- - `--progress plain` switches to human-readable progress lines.
77
- - background file locations are controlled by `workflow.background` in
78
- `settings.json`.
79
-
80
- ## Runtime Boundaries
81
-
82
- - Use Codex app-server over stdio as the production backend.
83
- - Keep direct provider credentials out of Codex child process environments.
84
- - Codex subagents run against the requested workflow cwd and have bounded
85
- read-only workspace tools for text file reads and directory listings.
86
- - Built-in `task` and `code-review` inject deterministic workspace context into
87
- planner-selected phase-wise parallel subagents.
88
- - Keep workflow execution local and command-owned; settings default to OS
89
- background execution so long runs can keep waiting while Codex does other
90
- work.
91
- - Keep `journalPath`, `journal.jsonl`, and journal contents out of CLI output.
92
- - Treat `.ultracode-for-codex` workflow state as sensitive local data.
93
- - Keep `resumeFromRunId` runtime-internal unless cross-process resume
94
- gets an explicit durable design.
95
- - Use `isolation: "worktree"` only inside a git repo with at least one commit;
96
- isolated worktrees are intentionally preserved for review, including clean
97
- worktrees.
98
-
99
- ## Packaging And Verification
100
-
101
- For source checkout changes, run the narrowest relevant check first, then a
102
- release-level check before handoff:
103
-
104
- ```bash
105
- npm test
106
- npm run test:e2e:ultracode-for-codex
107
- npm run test:all
108
- ```
21
+ Use Codex subagent tools for delegated work. If subagent tools are not visible,
22
+ search for the multi-agent tools first. If no subagent surface is available,
23
+ state that native parallel orchestration is unavailable in this session and
24
+ continue with the best single-context workflow.
109
25
 
110
- Build an installable artifact with:
26
+ Do not make the CLI process the default orchestrator for this skill. The npm
27
+ runtime remains available through `$ultracode-for-codex-cli`, but this command's
28
+ value is high-visibility orchestration in the main context.
111
29
 
112
- ```bash
113
- npm run pack:ultracode-for-codex
114
- ```
30
+ ## Native Workflow
115
31
 
116
- Check the npm publish payload with:
32
+ 1. Identify the user goal, scope, constraints, likely completion condition, and
33
+ whether the work is review, implementation, planning, verification, or mixed.
34
+ 2. Design only the next useful phase when later phases depend on earlier
35
+ results. A first plan may be partial.
36
+ 3. Before each phase starts, show a compact phase plan in the chat:
117
37
 
118
- ```bash
119
- npm run publish:dry-run
38
+ ```text
39
+ Phase Inspect - 3 agents
40
+ - Runtime contracts: verify the active execution path and failure semantics.
41
+ - UX/progress: check visibility, summaries, and user-facing wording.
42
+ - Tests/package: check coverage, package contents, and install behavior.
120
43
  ```
121
44
 
122
- Publish after npm login with:
123
-
124
- ```bash
125
- npm run publish:npm
45
+ 4. Spawn independent phase agents in parallel by default. Use a single agent
46
+ only when parallel work is risky, wasteful, or blocked by a strictly
47
+ sequential dependency.
48
+ 5. Keep subagent prompts concrete and bounded. Give each agent a distinct angle,
49
+ expected output shape, and file or responsibility boundary when relevant.
50
+ 6. While agents run, do non-overlapping main-context work such as deterministic
51
+ file inspection, test execution, or integration planning.
52
+ 7. As agents complete, report progress with a visual snapshot rather than a
53
+ dense sentence. Use the Default Live Snapshot golden shape from
54
+ `references/progress-visuals.md` by default.
55
+
56
+ ```text
57
+ Phase E2E Validate
58
+
59
+ + Native routing review done 34s
60
+ + CLI package review done 48s
61
+
62
+ > npm-exec-run-shim running 1/2 checks
63
+ > skill-copy-detection running 3/6 files
64
+
65
+ Agents 2 completed | 2 running
66
+ Checks 5 passed | 0 failed | 2 running
67
+ Elapsed 1m 12s
126
68
  ```
127
69
 
128
- When architecture, runtime boundaries, package exports, or release scope
129
- changes, update active docs such as `README.md`, `ULTRACODE_INSTALL.md`, and
130
- `IMPLEMENTATION_MAP.html`.
70
+ 8. Synthesize each phase before deciding the next phase. Preserve disagreement,
71
+ uncertainty, material risks, and exact evidence.
72
+ 9. After the final synthesis, include the Completion Impact Summary and
73
+ Plan-Style Result Summary golden shapes from `references/progress-visuals.md`,
74
+ followed by a short phase/agent summary.
75
+ 10. Recommend that the current session LLM critically re-check the final result
76
+ before the user relies on it, especially for code, security, release, or
77
+ architecture decisions.
78
+
79
+ ## Planning Heuristics
80
+
81
+ Default to phase-wise parallel execution. Useful patterns include:
82
+
83
+ - classify-and-act: classify request type, risk, or repo area before choosing
84
+ the phase shape;
85
+ - fan-out-and-synthesize: split independent lenses across parallel agents, then
86
+ merge evidence;
87
+ - adversarial verification: assign at least one agent to challenge correctness,
88
+ security, assumptions, or test adequacy;
89
+ - generate-and-filter: create candidate approaches or fixes, then select by
90
+ evidence and constraints;
91
+ - tournament: compare competing alternatives when the best path is unclear;
92
+ - loop-until-done: iterate repair and verification only when there is a clear
93
+ stop condition.
94
+
95
+ For code review, common parallel angles are runtime correctness,
96
+ security/capability boundaries, API/CLI contracts, persistence/retry/cancel
97
+ behavior, user-visible progress, package contents, and test coverage.
98
+
99
+ For implementation, split by disjoint write ownership where possible. Tell
100
+ subagents they are not alone in the codebase and must not revert unrelated or
101
+ parallel edits.
102
+
103
+ ## Output Contract
104
+
105
+ Keep progress visible but concise. Prefer stable visual summaries over prose-only
106
+ status sentences. Use `references/progress-visuals.md` for the golden examples.
107
+ The final answer should include:
108
+
109
+ - the completed result or findings;
110
+ - evidence and verification performed;
111
+ - a phase/agent summary;
112
+ - residual risk or unverified items;
113
+ - a critical re-check recommendation for the current session LLM when the result
114
+ will drive action.
@@ -1,4 +1,4 @@
1
1
  interface:
2
2
  display_name: "Ultracode for Codex"
3
- short_description: "Operate the Ultracode for Codex npm runtime safely."
4
- default_prompt: "Use the Ultracode for Codex npm package as the runtime and follow its local CLI boundaries."
3
+ short_description: "Run Codex-native phase-wise parallel orchestration."
4
+ default_prompt: "Use the Codex main context as the Ultracode orchestrator, plan adaptive phases, spawn focused parallel subagents, synthesize results, and keep progress visible."
@@ -0,0 +1,125 @@
1
+ # Progress Visuals
2
+
3
+ Use these golden examples for Codex-native Ultracode progress updates. The goal
4
+ is fast visual parsing in chat while staying portable across terminals and
5
+ renderers. Prefer ASCII symbols and short labels.
6
+
7
+ ## Default Live Snapshot
8
+
9
+ Use this for ordinary phase progress. It is inspired by test-runner summaries:
10
+ completed work is listed first, active work is visually distinct, and totals are
11
+ grouped at the bottom.
12
+
13
+ ```text
14
+ Phase E2E Validate
15
+
16
+ + Native routing review done 34s
17
+ + CLI package review done 48s
18
+
19
+ > npm-exec-run-shim running 1/2 checks
20
+ > skill-copy-detection running 3/6 files
21
+
22
+ Agents 2 completed | 2 running
23
+ Checks 5 passed | 0 failed | 2 running
24
+ Elapsed 1m 12s
25
+ ```
26
+
27
+ Use `+` for completed items, `>` for running items, `!` for blocked or failed
28
+ items, and `-` for queued items. Keep each row to one line when possible.
29
+
30
+ ## Dense Meter Snapshot
31
+
32
+ Use this when the work is count-heavy and the user needs ratios.
33
+
34
+ ```text
35
+ Progress Snapshot main c3734d8
36
+
37
+ Native routing ####################---- 5 / 6 checks
38
+ CLI package E2E ######################## passed
39
+ Skill install check ################-------- 4 / 6 files
40
+ Docs contract ############------------ 2 / 4 sections
41
+
42
+ Elapsed 1m 42s Next: npm exec run shim check
43
+ ```
44
+
45
+ Use fixed-width bars only when the denominator is meaningful. Do not invent a
46
+ percentage for semantic work that cannot be counted.
47
+
48
+ ## Long Async Timeline
49
+
50
+ Use this for long-running parallel work where event order matters.
51
+
52
+ ```text
53
+ #1 [plan] classify task shape
54
+ #1 DONE 0.8s
55
+
56
+ #2 [phase:inspect] spawn 2 review agents
57
+ #2 running 2 agents
58
+
59
+ #3 [agent:native-routing] verify skill split
60
+ #3 DONE 34.1s
61
+
62
+ #4 [agent:cli-package] verify installed E2E
63
+ #4 DONE 48.7s
64
+
65
+ #5 [synthesis] merge findings
66
+ #5 running
67
+ ```
68
+
69
+ Use this sparingly in chat. It is best when the user asks for detailed live
70
+ process visibility.
71
+
72
+ ## Completion Impact Summary
73
+
74
+ Use this in final or phase-completion reporting when files changed.
75
+
76
+ ```text
77
+ Change Impact
78
+
79
+ skills/ultracode-for-codex/SKILL.md | 200 +++++++++++---------
80
+ skills/ultracode-for-codex-cli/SKILL.md | 136 +++++++++++++
81
+ scripts/e2e-installed-ultracode-for-codex.mjs | 78 +++++++-
82
+ README.md | 62 ++++--
83
+
84
+ 8 files changed, 295 insertions(+), 201 deletions(-)
85
+ ```
86
+
87
+ Prefer real `git diff --stat` output when available.
88
+
89
+ ## Plan-Style Result Summary
90
+
91
+ Use this with the impact summary to explain what changed conceptually.
92
+
93
+ ```text
94
+ Execution Result
95
+
96
+ + add ultracode-for-codex-cli skill command
97
+ ~ update ultracode-for-codex native orchestration contract
98
+ ~ update installed E2E to cover npm exec run
99
+ = keep CLI runtime command surface
100
+
101
+ Result: 1 added, 2 updated, 1 kept
102
+ Risk: Codex skill reload cannot be forced inside the current session
103
+ ```
104
+
105
+ Use `+ add`, `~ update`, `= keep`, and `! risk` consistently.
106
+
107
+ ## Folded Detail Summary
108
+
109
+ Use this when the top-level result is enough and details should remain compact.
110
+
111
+ ```text
112
+ E2E Validation Summary
113
+
114
+ [passed] Native skill routing
115
+ [passed] CLI package runtime
116
+ [passed] npm exec run shim
117
+ [passed] CODEX_HOME skill copy
118
+
119
+ Details
120
+ native-routing-review 5 findings, 0 failures
121
+ cli-package-review 7 findings, 0 failures
122
+ npm test:all 35 tests passed
123
+ ```
124
+
125
+ This is best for final handoff or after multiple agents have returned.
@@ -0,0 +1,136 @@
1
+ ---
2
+ name: ultracode-for-codex-cli
3
+ description: Operate, package, validate, or update the Ultracode for Codex npm CLI runtime, including background jobs, attached runs, runtime boundaries, release tarballs, and Codex skill command files.
4
+ ---
5
+
6
+ # Ultracode for Codex CLI
7
+
8
+ ## Core Rule
9
+
10
+ Use this skill for the npm package and CLI runtime surface. Runtime authority for
11
+ this path lives in the `ultracode-for-codex` binary, tests, package exports,
12
+ journal layer, and workflow runtime code.
13
+
14
+ The default `$ultracode-for-codex` skill is Codex-native and keeps orchestration
15
+ in the main context. This `$ultracode-for-codex-cli` skill is for explicit CLI
16
+ runtime work: background execution, attached runs, package validation, release
17
+ preparation, installed E2E checks, runtime-boundary checks, and local workflow
18
+ artifacts.
19
+
20
+ Workflow execution through this path runs through the local CLI command.
21
+ Progress, cancellation, permission review, retry, and result projection stay in
22
+ that command process. `settings.json` defaults runs to OS background execution;
23
+ use that path for long Codex-launched CLI work so Codex can keep doing other
24
+ tasks and inspect the background job later. Attached runs stream stderr JSONL for
25
+ Codex-readable status, while stdout remains the final workflow result JSON.
26
+
27
+ ## Install And Run
28
+
29
+ Use the npm package for consumer installs.
30
+
31
+ ```bash
32
+ npm install --save-dev ultracode-for-codex
33
+ npm exec -- ultracode-for-codex --llm-guide
34
+ npm exec -- ultracode-for-codex run \
35
+ --accept-llm-guide=v1 \
36
+ --cwd /path/to/project \
37
+ --script-file .codex/workflows/review.js \
38
+ --args '{"prompt":"review the current change"}'
39
+ ```
40
+
41
+ For source-checkout validation before publish:
42
+
43
+ ```bash
44
+ npm run pack:ultracode-for-codex
45
+ npm install --save-dev ./artifacts/ultracode-for-codex-<version>.tgz
46
+ ```
47
+
48
+ CLI behavior:
49
+
50
+ - `--version` or `-v` prints the installed package version;
51
+ - default execution is `background`; stdout contains a launch record with
52
+ `jobId`, `pid`, `resultPath`, `progressPath`, `metadataPath`, and `pidPath`;
53
+ - background jobs can be inspected with `status`, waited with `wait`, read with
54
+ `logs` and `result`, and cancelled with `cancel`;
55
+ - background jobs can be enumerated with `jobs` or `list`, and exported without
56
+ deletion with `archive` or `export`;
57
+ - `wait --result`, `cancel --wait`, `logs --event <event>`, and `--plain`
58
+ provide focused foreground checks;
59
+ - attached execution is available with `--execution attached` when the caller
60
+ should stay connected until completion;
61
+ - attached progress prints to stderr as JSONL by default;
62
+ - attached final workflow result prints as JSON to stdout;
63
+ - JSONL records include `kind`, `version`, `event`, `status`, and `summary`,
64
+ with agent identity and label fields on agent records;
65
+ - built-in `task` and `code-review` emit `workflow.plan.ready` as a planning
66
+ snapshot, not a promise that every later phase is already known;
67
+ - `workflow.phase.planned` is emitted immediately before each phase starts and
68
+ carries that phase's current planned agent role labels;
69
+ - each `workflow.phase.started` record repeats the same role labels when the
70
+ phase begins;
71
+ - each `workflow.agent.completed` record includes phase progress, total known
72
+ agent progress, and elapsed time;
73
+ - after a completed run, `workflow.summary.ready` reports phase-level agent
74
+ counts and angles, then `workflow.review.recommended` asks the current
75
+ session LLM to critically re-check the final result before acting on it;
76
+ - `Ctrl-C` cancels the active attached workflow;
77
+ - `--retry-limit <n>` retries failed workflows inside the same process;
78
+ - `--timeout-ms 0` waits for completion, cancellation, or app-server exit;
79
+ positive values opt into a workflow deadline and per-agent silence budget,
80
+ and that budget is not divided by the retry budget.
81
+ - `--permission ask|allow|deny` handles project/user/plugin/scriptPath reviews.
82
+ - `--progress plain` switches to human-readable progress lines.
83
+ - background file locations are controlled by `workflow.background` in
84
+ `settings.json`.
85
+
86
+ ## Runtime Boundaries
87
+
88
+ - Use Codex app-server over stdio as the production backend.
89
+ - Keep direct provider credentials out of Codex child process environments.
90
+ - Codex subagents run against the requested workflow cwd and have bounded
91
+ read-only workspace tools for text file reads and directory listings.
92
+ - Built-in `task` and `code-review` inject deterministic workspace context into
93
+ planner-selected phase-wise parallel subagents.
94
+ - Keep workflow execution local and command-owned; settings default to OS
95
+ background execution so long runs can keep waiting while Codex does other
96
+ work.
97
+ - Keep `journalPath`, `journal.jsonl`, and journal contents out of CLI output.
98
+ - Treat `.ultracode-for-codex` workflow state as sensitive local data.
99
+ - Keep `resumeFromRunId` runtime-internal unless cross-process resume gets an
100
+ explicit durable design.
101
+ - Use `isolation: "worktree"` only inside a git repo with at least one commit;
102
+ isolated worktrees are intentionally preserved for review, including clean
103
+ worktrees.
104
+
105
+ ## Packaging And Verification
106
+
107
+ For source checkout changes, run the narrowest relevant check first, then a
108
+ release-level check before handoff:
109
+
110
+ ```bash
111
+ npm test
112
+ npm run test:e2e:ultracode-for-codex
113
+ npm run test:all
114
+ ```
115
+
116
+ Build an installable artifact with:
117
+
118
+ ```bash
119
+ npm run pack:ultracode-for-codex
120
+ ```
121
+
122
+ Check the npm publish payload with:
123
+
124
+ ```bash
125
+ npm run publish:dry-run
126
+ ```
127
+
128
+ Publish after npm login with:
129
+
130
+ ```bash
131
+ npm run publish:npm
132
+ ```
133
+
134
+ When architecture, runtime boundaries, package exports, or release scope
135
+ changes, update active docs such as `README.md`, `ULTRACODE_INSTALL.md`, and
136
+ `IMPLEMENTATION_MAP.html`.
@@ -0,0 +1,4 @@
1
+ interface:
2
+ display_name: "Ultracode for Codex CLI"
3
+ short_description: "Operate the Ultracode for Codex npm CLI runtime."
4
+ default_prompt: "Use the Ultracode for Codex CLI runtime for explicit package, background job, validation, and release work."