uctm 1.1.0 → 1.1.1
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.
- package/README.md +7 -19
- package/agents/en/agent-flow.md +106 -106
- package/agents/en/builder.md +5 -5
- package/agents/en/committer.md +7 -7
- package/agents/en/context-policy.md +1 -1
- package/agents/en/file-content-schema.md +224 -224
- package/agents/en/planner.md +5 -5
- package/agents/en/router.md +5 -5
- package/agents/en/scheduler.md +7 -7
- package/agents/en/shared-prompt-sections.md +1 -1
- package/agents/en/verifier.md +4 -4
- package/agents/en/work-activity-log.md +45 -45
- package/agents/en/xml-schema.md +109 -109
- package/agents/ko/agent-flow.md +106 -106
- package/agents/ko/builder.md +5 -5
- package/agents/ko/committer.md +7 -7
- package/agents/ko/context-policy.md +1 -1
- package/agents/ko/file-content-schema.md +224 -224
- package/agents/ko/planner.md +5 -5
- package/agents/ko/router.md +5 -5
- package/agents/ko/scheduler.md +7 -7
- package/agents/ko/shared-prompt-sections.md +1 -1
- package/agents/ko/verifier.md +4 -4
- package/agents/ko/work-activity-log.md +45 -45
- package/agents/ko/xml-schema.md +109 -109
- package/lib/constants.mjs +4 -4
- package/package.json +1 -1
package/README.md
CHANGED
|
@@ -60,24 +60,6 @@ uc-taskmanager was built to solve this. When you provide a requirement as input,
|
|
|
60
60
|
|
|
61
61
|
Each step runs in an **isolated subagent pipeline**, so your context stays clean and every decision is documented.
|
|
62
62
|
|
|
63
|
-
### The Bigger Picture
|
|
64
|
-
|
|
65
|
-
This agent is designed to work with an **SDD-based requirement management and automated development system** — a server application that links requirement management → automated development → plans and artifacts. The full system architecture is documented in [`docs/spec_SDD_with_ucagent_requirement.md`](docs/spec_SDD_with_ucagent_requirement.md). Use it as a reference to build your own system tailored to your needs.
|
|
66
|
-
|
|
67
|
-
### A Note on Language
|
|
68
|
-
|
|
69
|
-
Starting from v1.1.0, uctm supports **multi-language agent prompts**. Use `uctm init --lang en` for English or `--lang ko` for Korean. All 12 agent files are available in both languages with identical structure and functionality.
|
|
70
|
-
|
|
71
|
-
I'm Korean, so you'll find Korean in some internal documents (design specs, commit messages). Modern translation tools handle Korean well, so this shouldn't be a barrier.
|
|
72
|
-
|
|
73
|
-
### What This Took
|
|
74
|
-
|
|
75
|
-
I built this agent and the accompanying SDD requirement management system in roughly **one calendar month — about 2 weeks of actual development time** (8 hours/day). Ideas become reality. The technical and time barriers that once stopped you have fallen. This is the age of AI agents.
|
|
76
|
-
|
|
77
|
-
**Good luck. Build something great.**
|
|
78
|
-
|
|
79
|
-
This project is open source. Use it freely, share your ideas, and let's push this forward together.
|
|
80
|
-
|
|
81
63
|
---
|
|
82
64
|
|
|
83
65
|
Six subagents work across any project and any language, automatically handling **request routing → task decomposition → dependency management → code implementation → verification → commit**.
|
|
@@ -265,7 +247,7 @@ uctm update --lang en
|
|
|
265
247
|
```bash
|
|
266
248
|
git clone https://github.com/UCJung/uc-taskmanager-claude-agent.git /tmp/uc-tm
|
|
267
249
|
mkdir -p .claude/agents
|
|
268
|
-
cp /tmp/uc-tm/agents/*.md .claude/agents/
|
|
250
|
+
cp /tmp/uc-tm/agents/en/*.md .claude/agents/ # or agents/ko/ for Korean
|
|
269
251
|
rm -rf /tmp/uc-tm
|
|
270
252
|
git add .claude/agents/ && git commit -m "chore: add uc-taskmanager agents"
|
|
271
253
|
```
|
|
@@ -953,6 +935,12 @@ The dashboard is still available at `http://localhost:PORT` — it just won't au
|
|
|
953
935
|
|
|
954
936
|
---
|
|
955
937
|
|
|
938
|
+
## The Bigger Picture
|
|
939
|
+
|
|
940
|
+
This agent is designed to work with an **SDD-based requirement management and automated development system** — a server application that links requirement management → automated development → plans and artifacts. The full system architecture is documented in [`docs/spec_SDD_with_ucagent_requirement.md`](docs/spec_SDD_with_ucagent_requirement.md). Use it as a reference to build your own system tailored to your needs.
|
|
941
|
+
|
|
942
|
+
---
|
|
943
|
+
|
|
956
944
|
## License
|
|
957
945
|
|
|
958
946
|
GPL-3.0
|
package/agents/en/agent-flow.md
CHANGED
|
@@ -1,106 +1,106 @@
|
|
|
1
|
-
# Agent Flow — Main Claude Orchestration Guide
|
|
2
|
-
|
|
3
|
-
> **All agent invocations are performed by Main Claude.**
|
|
4
|
-
> Subagents only return results (dispatch XML or task-result XML) after completing their work.
|
|
5
|
-
> Main Claude receives the return values and invokes the next agent.
|
|
6
|
-
|
|
7
|
-
---
|
|
8
|
-
|
|
9
|
-
## Execution Mode Decision
|
|
10
|
-
|
|
11
|
-
```
|
|
12
|
-
[] tag detected → invoke router
|
|
13
|
-
│
|
|
14
|
-
Check router return value (execution-mode)
|
|
15
|
-
│
|
|
16
|
-
├─ direct → router handles everything (no additional invocations)
|
|
17
|
-
├─ pipeline → Execute § pipeline procedure
|
|
18
|
-
└─ full → Execute § full procedure
|
|
19
|
-
```
|
|
20
|
-
|
|
21
|
-
---
|
|
22
|
-
|
|
23
|
-
## direct Mode
|
|
24
|
-
|
|
25
|
-
Router handles everything on its own. No additional invocations by Main Claude.
|
|
26
|
-
|
|
27
|
-
---
|
|
28
|
-
|
|
29
|
-
## pipeline Mode
|
|
30
|
-
|
|
31
|
-
```
|
|
32
|
-
1. Invoke router → creates PLAN.md + TASK-00.md + returns builder dispatch XML
|
|
33
|
-
2. Invoke builder (dispatch XML as prompt)
|
|
34
|
-
3. Invoke verifier (builder result as prompt)
|
|
35
|
-
4. Invoke committer (verifier result as prompt)
|
|
36
|
-
```
|
|
37
|
-
|
|
38
|
-
---
|
|
39
|
-
|
|
40
|
-
## full Mode
|
|
41
|
-
|
|
42
|
-
```
|
|
43
|
-
1. Invoke router → creates WORK directory + returns planner dispatch XML
|
|
44
|
-
2. Invoke planner (dispatch XML as prompt) → creates PLAN.md + TASK files
|
|
45
|
-
3. Invoke scheduler → DAG analysis + READY TASK + returns builder dispatch XML
|
|
46
|
-
4. Invoke builder (dispatch XML as prompt) → implementation
|
|
47
|
-
5. Invoke verifier (builder result as prompt) → verification
|
|
48
|
-
6. Invoke committer (verifier result as prompt) → commit
|
|
49
|
-
7. If incomplete TASKs remain, return to step 3
|
|
50
|
-
```
|
|
51
|
-
|
|
52
|
-
Parallel execution: If scheduler returns multiple READY TASKs, invoke builders concurrently.
|
|
53
|
-
|
|
54
|
-
---
|
|
55
|
-
|
|
56
|
-
## Agent Role Summary
|
|
57
|
-
|
|
58
|
-
| Agent | Return Value | Invoked By |
|
|
59
|
-
|-------|-------------|------------|
|
|
60
|
-
| router | execution-mode + dispatch XML | Main Claude |
|
|
61
|
-
| planner | PLAN.md/TASK file creation completion report | Main Claude |
|
|
62
|
-
| scheduler | READY TASK + dispatch XML | Main Claude |
|
|
63
|
-
| builder | task-result XML (includes context-handoff) | Main Claude |
|
|
64
|
-
| verifier | task-result XML | Main Claude |
|
|
65
|
-
| committer | task-result XML + commit hash | Main Claude |
|
|
66
|
-
|
|
67
|
-
---
|
|
68
|
-
|
|
69
|
-
## Bash CLI Execution (Server Automation)
|
|
70
|
-
|
|
71
|
-
Method to run pipeline independently without an interactive session. `claude -p` acts as Main Claude.
|
|
72
|
-
|
|
73
|
-
```bash
|
|
74
|
-
env -u CLAUDECODE -u ANTHROPIC_API_KEY claude -p \
|
|
75
|
-
"[new-work] {task description}" \
|
|
76
|
-
--dangerously-skip-permissions \
|
|
77
|
-
--output-format stream-json \
|
|
78
|
-
--verbose \
|
|
79
|
-
2>&1 | tee /tmp/pipeline.log
|
|
80
|
-
```
|
|
81
|
-
|
|
82
|
-
| Option | Purpose |
|
|
83
|
-
|--------|---------|
|
|
84
|
-
| `env -u CLAUDECODE` | Bypass nested execution blocking |
|
|
85
|
-
| `env -u ANTHROPIC_API_KEY` | Use subscription auth (Max) instead of API key |
|
|
86
|
-
| `--dangerously-skip-permissions` | Skip permission prompts for unattended execution |
|
|
87
|
-
| `--output-format stream-json --verbose` | Streaming for real-time monitoring |
|
|
88
|
-
|
|
89
|
-
Resume interrupted pipeline:
|
|
90
|
-
```bash
|
|
91
|
-
env -u CLAUDECODE -u ANTHROPIC_API_KEY claude -p \
|
|
92
|
-
"Resume WORK-XX pipeline." \
|
|
93
|
-
--dangerously-skip-permissions
|
|
94
|
-
```
|
|
95
|
-
|
|
96
|
-
Verification result (WORK-24): `claude -p` → 9 Task tool invocations → full auto-completion of router/planner/scheduler/builder/verifier/committer confirmed.
|
|
97
|
-
|
|
98
|
-
---
|
|
99
|
-
|
|
100
|
-
## Context Transfer (Sliding Window)
|
|
101
|
-
|
|
102
|
-
| Distance | Level | Content |
|
|
103
|
-
|----------|-------|---------|
|
|
104
|
-
| Immediate predecessor | FULL | what + why + caution + incomplete |
|
|
105
|
-
| 2 steps back | SUMMARY | what only, 1-2 lines |
|
|
106
|
-
| 3+ steps back | DROP | Not transmitted |
|
|
1
|
+
# Agent Flow — Main Claude Orchestration Guide
|
|
2
|
+
|
|
3
|
+
> **All agent invocations are performed by Main Claude.**
|
|
4
|
+
> Subagents only return results (dispatch XML or task-result XML) after completing their work.
|
|
5
|
+
> Main Claude receives the return values and invokes the next agent.
|
|
6
|
+
|
|
7
|
+
---
|
|
8
|
+
|
|
9
|
+
## Execution Mode Decision
|
|
10
|
+
|
|
11
|
+
```
|
|
12
|
+
[] tag detected → invoke router
|
|
13
|
+
│
|
|
14
|
+
Check router return value (execution-mode)
|
|
15
|
+
│
|
|
16
|
+
├─ direct → router handles everything (no additional invocations)
|
|
17
|
+
├─ pipeline → Execute § pipeline procedure
|
|
18
|
+
└─ full → Execute § full procedure
|
|
19
|
+
```
|
|
20
|
+
|
|
21
|
+
---
|
|
22
|
+
|
|
23
|
+
## direct Mode
|
|
24
|
+
|
|
25
|
+
Router handles everything on its own. No additional invocations by Main Claude.
|
|
26
|
+
|
|
27
|
+
---
|
|
28
|
+
|
|
29
|
+
## pipeline Mode
|
|
30
|
+
|
|
31
|
+
```
|
|
32
|
+
1. Invoke router → creates PLAN.md + TASK-00.md + returns builder dispatch XML
|
|
33
|
+
2. Invoke builder (dispatch XML as prompt)
|
|
34
|
+
3. Invoke verifier (builder result as prompt)
|
|
35
|
+
4. Invoke committer (verifier result as prompt)
|
|
36
|
+
```
|
|
37
|
+
|
|
38
|
+
---
|
|
39
|
+
|
|
40
|
+
## full Mode
|
|
41
|
+
|
|
42
|
+
```
|
|
43
|
+
1. Invoke router → creates WORK directory + returns planner dispatch XML
|
|
44
|
+
2. Invoke planner (dispatch XML as prompt) → creates PLAN.md + TASK files
|
|
45
|
+
3. Invoke scheduler → DAG analysis + READY TASK + returns builder dispatch XML
|
|
46
|
+
4. Invoke builder (dispatch XML as prompt) → implementation
|
|
47
|
+
5. Invoke verifier (builder result as prompt) → verification
|
|
48
|
+
6. Invoke committer (verifier result as prompt) → commit
|
|
49
|
+
7. If incomplete TASKs remain, return to step 3
|
|
50
|
+
```
|
|
51
|
+
|
|
52
|
+
Parallel execution: If scheduler returns multiple READY TASKs, invoke builders concurrently.
|
|
53
|
+
|
|
54
|
+
---
|
|
55
|
+
|
|
56
|
+
## Agent Role Summary
|
|
57
|
+
|
|
58
|
+
| Agent | Return Value | Invoked By |
|
|
59
|
+
|-------|-------------|------------|
|
|
60
|
+
| router | execution-mode + dispatch XML | Main Claude |
|
|
61
|
+
| planner | PLAN.md/TASK file creation completion report | Main Claude |
|
|
62
|
+
| scheduler | READY TASK + dispatch XML | Main Claude |
|
|
63
|
+
| builder | task-result XML (includes context-handoff) | Main Claude |
|
|
64
|
+
| verifier | task-result XML | Main Claude |
|
|
65
|
+
| committer | task-result XML + commit hash | Main Claude |
|
|
66
|
+
|
|
67
|
+
---
|
|
68
|
+
|
|
69
|
+
## Bash CLI Execution (Server Automation)
|
|
70
|
+
|
|
71
|
+
Method to run pipeline independently without an interactive session. `claude -p` acts as Main Claude.
|
|
72
|
+
|
|
73
|
+
```bash
|
|
74
|
+
env -u CLAUDECODE -u ANTHROPIC_API_KEY claude -p \
|
|
75
|
+
"[new-work] {task description}" \
|
|
76
|
+
--dangerously-skip-permissions \
|
|
77
|
+
--output-format stream-json \
|
|
78
|
+
--verbose \
|
|
79
|
+
2>&1 | tee /tmp/pipeline.log
|
|
80
|
+
```
|
|
81
|
+
|
|
82
|
+
| Option | Purpose |
|
|
83
|
+
|--------|---------|
|
|
84
|
+
| `env -u CLAUDECODE` | Bypass nested execution blocking |
|
|
85
|
+
| `env -u ANTHROPIC_API_KEY` | Use subscription auth (Max) instead of API key |
|
|
86
|
+
| `--dangerously-skip-permissions` | Skip permission prompts for unattended execution |
|
|
87
|
+
| `--output-format stream-json --verbose` | Streaming for real-time monitoring |
|
|
88
|
+
|
|
89
|
+
Resume interrupted pipeline:
|
|
90
|
+
```bash
|
|
91
|
+
env -u CLAUDECODE -u ANTHROPIC_API_KEY claude -p \
|
|
92
|
+
"Resume WORK-XX pipeline." \
|
|
93
|
+
--dangerously-skip-permissions
|
|
94
|
+
```
|
|
95
|
+
|
|
96
|
+
Verification result (WORK-24): `claude -p` → 9 Task tool invocations → full auto-completion of router/planner/scheduler/builder/verifier/committer confirmed.
|
|
97
|
+
|
|
98
|
+
---
|
|
99
|
+
|
|
100
|
+
## Context Transfer (Sliding Window)
|
|
101
|
+
|
|
102
|
+
| Distance | Level | Content |
|
|
103
|
+
|----------|-------|---------|
|
|
104
|
+
| Immediate predecessor | FULL | what + why + caution + incomplete |
|
|
105
|
+
| 2 steps back | SUMMARY | what only, 1-2 lines |
|
|
106
|
+
| 3+ steps back | DROP | Not transmitted |
|
package/agents/en/builder.md
CHANGED
|
@@ -35,11 +35,11 @@ You are the **Builder** — the implementation agent that receives a TASK specif
|
|
|
35
35
|
|
|
36
36
|
| File | Purpose |
|
|
37
37
|
|------|---------|
|
|
38
|
-
|
|
|
39
|
-
|
|
|
40
|
-
|
|
|
41
|
-
|
|
|
42
|
-
|
|
|
38
|
+
| `.claude/agents/file-content-schema.md` | File format schema |
|
|
39
|
+
| `.claude/agents/shared-prompt-sections.md` | Common rules (TASK ID, PLAN.md 7 fields, WORK-LIST) |
|
|
40
|
+
| `.claude/agents/xml-schema.md` | XML communication format |
|
|
41
|
+
| `.claude/agents/context-policy.md` | Sliding window rules |
|
|
42
|
+
| `.claude/agents/work-activity-log.md` | Activity Log rules (log_work function, STAGE table) |
|
|
43
43
|
|
|
44
44
|
### 3-2. XML Input Parsing
|
|
45
45
|
|
package/agents/en/committer.md
CHANGED
|
@@ -35,11 +35,11 @@ You are the **Committer** — the agent that generates the result report for a v
|
|
|
35
35
|
|
|
36
36
|
| File | Purpose |
|
|
37
37
|
|------|---------|
|
|
38
|
-
|
|
|
39
|
-
|
|
|
40
|
-
|
|
|
41
|
-
|
|
|
42
|
-
|
|
|
38
|
+
| `.claude/agents/file-content-schema.md` | File format schema |
|
|
39
|
+
| `.claude/agents/shared-prompt-sections.md` | Common rules |
|
|
40
|
+
| `.claude/agents/xml-schema.md` | XML communication format |
|
|
41
|
+
| `.claude/agents/context-policy.md` | Sliding window rules |
|
|
42
|
+
| `.claude/agents/work-activity-log.md` | Activity Log rules (log_work function, STAGE table) |
|
|
43
43
|
|
|
44
44
|
### 3-2. XML Input Parsing
|
|
45
45
|
|
|
@@ -66,7 +66,7 @@ On gate failure:
|
|
|
66
66
|
|
|
67
67
|
### 3-4. Result Report Generation
|
|
68
68
|
|
|
69
|
-
→ see
|
|
69
|
+
→ see `.claude/agents/file-content-schema.md` § 4 (format + language-specific section headers)
|
|
70
70
|
|
|
71
71
|
Create `works/{WORK_ID}/TASK-XX_result.md`.
|
|
72
72
|
- builder context-handoff `what` → "Builder Context" section
|
|
@@ -156,7 +156,7 @@ Committer-specific additional fields:
|
|
|
156
156
|
```
|
|
157
157
|
|
|
158
158
|
Do not change WORK-LIST.md to COMPLETED — changed only at git push time.
|
|
159
|
-
→ see
|
|
159
|
+
→ see `.claude/agents/shared-prompt-sections.md` § 8
|
|
160
160
|
|
|
161
161
|
---
|
|
162
162
|
|
|
@@ -56,7 +56,7 @@ Processing:
|
|
|
56
56
|
2. Gate passed → write result.md + git commit
|
|
57
57
|
3. Gate failed → return FAIL (triggers scheduler retry)
|
|
58
58
|
|
|
59
|
-
Output: →
|
|
59
|
+
Output: → `.claude/agents/file-content-schema.md` § 4 reference
|
|
60
60
|
|
|
61
61
|
## Inter-TASK Dependency Transfer
|
|
62
62
|
|