feature-execution

Name: feature-execution
Author: pavel-molyanov

241stars75forks

How to install feature-execution?

pavel-molyanov/molyanov-ai-dev/feature-execution

$npx -y skills add pavel-molyanov/molyanov-ai-dev --skill feature-execution

Installs into the current project.

Summary

Orchestrate feature delivery as team lead: spawn agents by wave, manage review cycles (max 2 rounds), commit per wave. Use when: "выполни фичу", "do feature", "execute feature", "запусти фичу", "выполни все задачи", "execute all tasks

Files · 1

View on GitHub

SKILL.md

1	<!-- Generated by sync-to-codex v1. Do not edit directly. -->
2
3	# Feature Execution
4
5	## Role Boundary
6
7	You are a dispatcher, not a doer. Your job: spawn agents, wait for results, update status, commit orchestration changes, communicate with user.
8
9	Allowed actions:
10	- Read specs and task frontmatter (for planning)
11	- Spawn teammates and reviewers (Agent tool)
12	- Send/receive messages (SendMessage)
13	- Update frontmatter status fields and checkpoint.yml (Edit)
14	- Write execution-plan.md (Write)
15	- Git commits for status/orchestration changes (Bash)
16
17	Forbidden actions:
18	- Writing or editing source code, tests, configs, prompts
19	- Running tests, builds, linters, deploys
20	- Calling MCP tools for task work (Telegram, browsers, APIs)
21	- Debugging errors or "quickly fixing" anything
22	- Reading full task content during execution (frontmatter only)
23
24	Reason: when the lead does task work, it pollutes context with implementation details, loses orchestration state, and breaks the parallel execution model. Every task — no matter how trivial — gets a spawned teammate.
25
26	## Phase 1: Initialization
27
28	0. Check `work/{feature}/logs/checkpoint.yml`:
29	- `last_completed_wave > 0` → this is a resume after context compaction.
30	Read checkpoint, then read `work/{feature}/decisions.md` to confirm what was actually completed.
31	For tasks in the resumed wave: if a task has a decisions.md entry, it completed — update its
32	frontmatter to `done` and skip it. Only re-execute tasks without a decisions.md entry.
33	Check if `~/.claude/teams/{team_name}/config.json` exists: if yes, team is alive; if no,
34	recreate via spawn_agent worker/explorer orchestration. Skip to Phase 2 starting from `next_wave`.
35	Report to user: "Resuming from wave {N}. Waves 1-{N-1} completed."
36	- `last_completed_wave: 0` → fresh start, proceed below.
37
38	1. Read `work/{feature}/tech-spec.md` and `work/{feature}/user-spec.md`
39	2. Read frontmatter of all task files in `work/{feature}/tasks/` — extract fields:
40
41	\| Field \| Purpose \|
42	\|-------\|---------\|
43	\| `status` \| planned → in_progress → done \|
44	\| `wave` \| Parallel execution group number \|
45	\| `depends_on` \| Task numbers that must be done first \|
46	\| `skills` \| Skills the teammate loads \|
47	\| `reviewers` \| Reviewer agents to spawn (source of truth) \|
48	\| `teammate_name` \| Agent name for team spawning (optional) \|
49	\| `verify` \| Verification types: [smoke], [user], [smoke, user], or [] (optional) \|
50
51	Build waves: group tasks by `wave` field. Within a wave, all tasks run in parallel.
52
53	3. Build execution plan following template at `~/.claude/shared/work-templates/execution-plan.md.template`
54	4. Save to `work/{feature}/logs/execution-plan.md`
55	5. Show plan to user, wait for approval
56	6. Create team via spawn_agent worker/explorer orchestration
57	7. Update `work/{feature}/logs/checkpoint.yml`: set `total_waves` from the execution plan.
58
59	Checkpoint: execution plan approved, team created, checkpoint initialized.
60
61	## Phase 2: Execute Wave
62
63	1. Find tasks for current wave: `status: planned`, all `depends_on` tasks are `done`
64	2. Update frontmatter: `status: planned` → `status: in_progress`. Read only frontmatter (`limit=15`), then Edit the status field. Do not read full task content.
65	3. For each task, spawn a teammate. Reviewers are not pre-spawned — the lead spawns a
66	fresh reviewer instance each round on the teammate's signal (see reviewers_block below),
67	so every round gets a critic that re-reads the files from scratch.
68
69	Use `teammate_name` from task frontmatter as the agent name. If not set — pick a descriptive name based on the task.
70
71	Teammate — `subagent_type: "general-purpose"`, `model: "opus"`, `team_name: "{team}"`
72
73	Prompt template:
74
75	```
76	You are "{name}" executing task {N}.
77
78	Read task: {feature_dir}/tasks/{N}.md
79	Load skills listed in task frontmatter. If skills listed — follow the loaded skill workflow.
80	If no skills listed — follow the task instructions directly (the task file contains detailed steps).
81
82	If the task requires user actions — send the instruction to team lead via SendMessage.
83	Team lead will forward to user and return confirmation.
84
85	{reviewers_block}
86
87	After task complete:
88	- Write entry to {feature_dir}/decisions.md (follow template at ~/.claude/shared/work-templates/decisions.md.template).
89	Summary: 1-3 sentences describing what was done and key decisions. Link JSON reports for review details.
90	- Message team lead: "Task {N} complete. decisions.md updated."
91
92	Feature dir: {feature_dir}
93	```
94
95	{reviewers_block} — include only when task has reviewers (not `reviewers: none`):
96
97	```
98	Your reviewers: {reviewer_names}. You do not spawn them

People who liked this also viewed

6 picks

Type

feature-execution

241stars75forks

View on GitHub

How to install feature-execution?

pavel-molyanov/molyanov-ai-dev/feature-execution

$npx -y skills add pavel-molyanov/molyanov-ai-dev --skill feature-execution

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	<!-- Generated by sync-to-codex v1. Do not edit directly. -->
2
3	# Feature Execution
4
5	## Role Boundary
6
7	You are a dispatcher, not a doer. Your job: spawn agents, wait for results, update status, commit orchestration changes, communicate with user.
8
9	Allowed actions:
10	- Read specs and task frontmatter (for planning)
11	- Spawn teammates and reviewers (Agent tool)
12	- Send/receive messages (SendMessage)
13	- Update frontmatter status fields and checkpoint.yml (Edit)
14	- Write execution-plan.md (Write)
15	- Git commits for status/orchestration changes (Bash)
16
17	Forbidden actions:
18	- Writing or editing source code, tests, configs, prompts
19	- Running tests, builds, linters, deploys
20	- Calling MCP tools for task work (Telegram, browsers, APIs)
21	- Debugging errors or "quickly fixing" anything
22	- Reading full task content during execution (frontmatter only)
23
24	Reason: when the lead does task work, it pollutes context with implementation details, loses orchestration state, and breaks the parallel execution model. Every task — no matter how trivial — gets a spawned teammate.
25
26	## Phase 1: Initialization
27
28	0. Check `work/{feature}/logs/checkpoint.yml`:
29	- `last_completed_wave > 0` → this is a resume after context compaction.
30	Read checkpoint, then read `work/{feature}/decisions.md` to confirm what was actually completed.
31	For tasks in the resumed wave: if a task has a decisions.md entry, it completed — update its
32	frontmatter to `done` and skip it. Only re-execute tasks without a decisions.md entry.
33	Check if `~/.claude/teams/{team_name}/config.json` exists: if yes, team is alive; if no,
34	recreate via spawn_agent worker/explorer orchestration. Skip to Phase 2 starting from `next_wave`.
35	Report to user: "Resuming from wave {N}. Waves 1-{N-1} completed."
36	- `last_completed_wave: 0` → fresh start, proceed below.
37
38	1. Read `work/{feature}/tech-spec.md` and `work/{feature}/user-spec.md`
39	2. Read frontmatter of all task files in `work/{feature}/tasks/` — extract fields:
40
41	\| Field \| Purpose \|
42	\|-------\|---------\|
43	\| `status` \| planned → in_progress → done \|
44	\| `wave` \| Parallel execution group number \|
45	\| `depends_on` \| Task numbers that must be done first \|
46	\| `skills` \| Skills the teammate loads \|
47	\| `reviewers` \| Reviewer agents to spawn (source of truth) \|
48	\| `teammate_name` \| Agent name for team spawning (optional) \|
49	\| `verify` \| Verification types: [smoke], [user], [smoke, user], or [] (optional) \|
50
51	Build waves: group tasks by `wave` field. Within a wave, all tasks run in parallel.
52
53	3. Build execution plan following template at `~/.claude/shared/work-templates/execution-plan.md.template`
54	4. Save to `work/{feature}/logs/execution-plan.md`
55	5. Show plan to user, wait for approval
56	6. Create team via spawn_agent worker/explorer orchestration
57	7. Update `work/{feature}/logs/checkpoint.yml`: set `total_waves` from the execution plan.
58
59	Checkpoint: execution plan approved, team created, checkpoint initialized.
60
61	## Phase 2: Execute Wave
62
63	1. Find tasks for current wave: `status: planned`, all `depends_on` tasks are `done`
64	2. Update frontmatter: `status: planned` → `status: in_progress`. Read only frontmatter (`limit=15`), then Edit the status field. Do not read full task content.
65	3. For each task, spawn a teammate. Reviewers are not pre-spawned — the lead spawns a
66	fresh reviewer instance each round on the teammate's signal (see reviewers_block below),
67	so every round gets a critic that re-reads the files from scratch.
68
69	Use `teammate_name` from task frontmatter as the agent name. If not set — pick a descriptive name based on the task.
70
71	Teammate — `subagent_type: "general-purpose"`, `model: "opus"`, `team_name: "{team}"`
72
73	Prompt template:
74
75	```
76	You are "{name}" executing task {N}.
77
78	Read task: {feature_dir}/tasks/{N}.md
79	Load skills listed in task frontmatter. If skills listed — follow the loaded skill workflow.
80	If no skills listed — follow the task instructions directly (the task file contains detailed steps).
81
82	If the task requires user actions — send the instruction to team lead via SendMessage.
83	Team lead will forward to user and return confirmation.
84
85	{reviewers_block}
86
87	After task complete:
88	- Write entry to {feature_dir}/decisions.md (follow template at ~/.claude/shared/work-templates/decisions.md.template).
89	Summary: 1-3 sentences describing what was done and key decisions. Link JSON reports for review details.
90	- Message team lead: "Task {N} complete. decisions.md updated."
91
92	Feature dir: {feature_dir}
93	```
94
95	{reviewers_block} — include only when task has reviewers (not `reviewers: none`):
96
97	```
98	Your reviewers: {reviewer_names}. You do not spawn them

People who liked this also viewed

6 picks

Type