delegation-strategy

Name: delegation-strategy
Author: omnigentx

31stars13forks

How to install delegation-strategy?

omnigentx/jarvis/delegation-strategy

$npx -y skills add omnigentx/jarvis --skill delegation-strategy

Installs into the current project.

Summary

Guides task delegation to specialized agents and spawn tools. Determines when to self-handle vs delegate, selects appropriate MCP servers and skills for spawned agents. Use when receiving research tasks, web access needs, or complex multi-step requests.

Files · 1

View on GitHub

SKILL.md

1	# Delegation Strategy
2
3	## Core Rule
4
5	Delegate ONLY when the task needs a specialized tool/data source (web, email/calendar, IoT, finance, media).
6	Reply directly for chitchat, venting, advice, opinions, or general Q&A — these are NOT delegation targets.
7	Among delegations: use existing agents first, spawn new only when no existing agent fits.
8
9	## Decision Flow
10
11	1. Ambiguous / underspecified? → Ask the user to clarify the expected output FIRST. Then plan and pick: single agent, multiple agents, or an existing team template (`list_team_templates_tool` → `spawn_team_tool`).
12	2. Existing agent matches? → Use `agent__<Name>` tool directly
13	3. No match, short task (<2 min)? → `spawn_and_run_isolated`
14	4. No match, long task? → `spawn_and_run_background`
15
16	## Existing Agents
17
18	\| Agent \| Domain \|
19	\|-------\|--------\|
20	\| ResearchAgent \| Web search, news, information synthesis \|
21	\| FinanceAgent \| Stock prices, gold, crypto, market analysis \|
22	\| PersonalAgent \| Email, calendar events, reminders (cron) — task management only, NOT casual conversation \|
23	\| CrawlStoriesAgent \| Story crawling from web \|
24	\| IoTAgent \| Smart home device control \|
25	\| MusicAgent \| Music search and playback \|
26	\| AudioReaderAgent \| Story audio playback \|
27
28	> Note: PersonalAgent uses the cron scheduler for reminders, NOT Google Calendar.
29	> Google Calendar is only for real events / meetings.
30
31	## Spawn Tool Quick Reference
32
33	```
34	spawn_and_run_isolated(
35	task="...", # Specific task description
36	instruction="...", # Agent role and behavior
37	servers="serpapi, scrapling-server", # Comma-separated
38	skills="research, scrape-web", # Comma-separated
39	role="researcher", # Label for tracking
40	timeout_seconds=90 # Default 120
41	)
42	```
43
44	For background: same params, use `spawn_and_run_background`, default timeout 600s.
45	Check status with `check_spawn_status(run_id)`.
46
47	## Choosing Servers and Skills
48
49	See [SERVERS.md](SERVERS.md) for MCP server catalog and selection rules.
50	See [SKILLS.md](SKILLS.md) for skill catalog and pairing guidance.
51
52	## Anti-patterns
53
54	- ❌ Jarvis self-researching (no direct web server access)
55	- ❌ Delegating chitchat / venting / advice to PersonalAgent — reply directly instead
56	- ❌ Spawning without servers (agent can't do anything)
57	- ❌ Adding `chrome-devtools` for simple search (use `serpapi`)
58	- ❌ Empty or vague instruction (always specify role + output format)
59	- ❌ Missing context (pass relevant conversation context to spawned agent)

People who liked this also viewed

6 picks

Type

delegation-strategy

31stars13forks

View on GitHub

How to install delegation-strategy?

omnigentx/jarvis/delegation-strategy

$npx -y skills add omnigentx/jarvis --skill delegation-strategy

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# Delegation Strategy
2
3	## Core Rule
4
5	Delegate ONLY when the task needs a specialized tool/data source (web, email/calendar, IoT, finance, media).
6	Reply directly for chitchat, venting, advice, opinions, or general Q&A — these are NOT delegation targets.
7	Among delegations: use existing agents first, spawn new only when no existing agent fits.
8
9	## Decision Flow
10
11	1. Ambiguous / underspecified? → Ask the user to clarify the expected output FIRST. Then plan and pick: single agent, multiple agents, or an existing team template (`list_team_templates_tool` → `spawn_team_tool`).
12	2. Existing agent matches? → Use `agent__<Name>` tool directly
13	3. No match, short task (<2 min)? → `spawn_and_run_isolated`
14	4. No match, long task? → `spawn_and_run_background`
15
16	## Existing Agents
17
18	\| Agent \| Domain \|
19	\|-------\|--------\|
20	\| ResearchAgent \| Web search, news, information synthesis \|
21	\| FinanceAgent \| Stock prices, gold, crypto, market analysis \|
22	\| PersonalAgent \| Email, calendar events, reminders (cron) — task management only, NOT casual conversation \|
23	\| CrawlStoriesAgent \| Story crawling from web \|
24	\| IoTAgent \| Smart home device control \|
25	\| MusicAgent \| Music search and playback \|
26	\| AudioReaderAgent \| Story audio playback \|
27
28	> Note: PersonalAgent uses the cron scheduler for reminders, NOT Google Calendar.
29	> Google Calendar is only for real events / meetings.
30
31	## Spawn Tool Quick Reference
32
33	```
34	spawn_and_run_isolated(
35	task="...", # Specific task description
36	instruction="...", # Agent role and behavior
37	servers="serpapi, scrapling-server", # Comma-separated
38	skills="research, scrape-web", # Comma-separated
39	role="researcher", # Label for tracking
40	timeout_seconds=90 # Default 120
41	)
42	```
43
44	For background: same params, use `spawn_and_run_background`, default timeout 600s.
45	Check status with `check_spawn_status(run_id)`.
46
47	## Choosing Servers and Skills
48
49	See [SERVERS.md](SERVERS.md) for MCP server catalog and selection rules.
50	See [SKILLS.md](SKILLS.md) for skill catalog and pairing guidance.
51
52	## Anti-patterns
53
54	- ❌ Jarvis self-researching (no direct web server access)
55	- ❌ Delegating chitchat / venting / advice to PersonalAgent — reply directly instead
56	- ❌ Spawning without servers (agent can't do anything)
57	- ❌ Adding `chrome-devtools` for simple search (use `serpapi`)
58	- ❌ Empty or vague instruction (always specify role + output format)
59	- ❌ Missing context (pass relevant conversation context to spawned agent)

People who liked this also viewed

6 picks

Type