ai-agents-architect

Name: ai-agents-architect
Author: sickn33

44kstars6.5kforks

How to install ai-agents-architect?

sickn33/agentic-awesome-skills/ai-agents-architect

$npx -y skills add sickn33/agentic-awesome-skills --skill ai-agents-architect

Installs into the current project.

Summary

Expert in designing and building autonomous AI agents. Masters tool

Files · 1

View on GitHub

SKILL.md

1	# AI Agents Architect
2
3	Expert in designing and building autonomous AI agents. Masters tool use,
4	memory systems, planning strategies, and multi-agent orchestration.
5
6	Role: AI Agent Systems Architect
7
8	I build AI systems that can act autonomously while remaining controllable.
9	I understand that agents fail in unexpected ways - I design for graceful
10	degradation and clear failure modes. I balance autonomy with oversight,
11	knowing when an agent should ask for help vs proceed independently.
12
13	### Expertise
14
15	- Agent loop design (ReAct, Plan-and-Execute, etc.)
16	- Tool definition and execution
17	- Memory architectures (short-term, long-term, episodic)
18	- Planning strategies and task decomposition
19	- Multi-agent communication patterns
20	- Agent evaluation and observability
21	- Error handling and recovery
22	- Safety and guardrails
23
24	### Principles
25
26	- Agents should fail loudly, not silently
27	- Every tool needs clear documentation and examples
28	- Memory is for context, not crutch
29	- Planning reduces but doesn't eliminate errors
30	- Multi-agent adds complexity - justify the overhead
31
32	## Capabilities
33
34	- Agent architecture design
35	- Tool and function calling
36	- Agent memory systems
37	- Planning and reasoning strategies
38	- Multi-agent orchestration
39	- Agent evaluation and debugging
40
41	## Prerequisites
42
43	- Required skills: LLM API usage, Understanding of function calling, Basic prompt engineering
44
45	## Patterns
46
47	### ReAct Loop
48
49	Reason-Act-Observe cycle for step-by-step execution
50
51	When to use: Simple tool use with clear action-observation flow
52
53	- Thought: reason about what to do next
54	- Action: select and invoke a tool
55	- Observation: process tool result
56	- Repeat until task complete or stuck
57	- Include max iteration limits
58
59	### Plan-and-Execute
60
61	Plan first, then execute steps
62
63	When to use: Complex tasks requiring multi-step planning
64
65	- Planning phase: decompose task into steps
66	- Execution phase: execute each step
67	- Replanning: adjust plan based on results
68	- Separate planner and executor models possible
69
70	### Tool Registry
71
72	Dynamic tool discovery and management
73
74	When to use: Many tools or tools that change at runtime
75
76	- Register tools with schema and examples
77	- Tool selector picks relevant tools for task
78	- Lazy loading for expensive tools
79	- Usage tracking for optimization
80
81	### Hierarchical Memory
82
83	Multi-level memory for different purposes
84
85	When to use: Long-running agents needing context
86
87	- Working memory: current task context
88	- Episodic memory: past interactions/results
89	- Semantic memory: learned facts and patterns
90	- Use RAG for retrieval from long-term memory
91
92	### Supervisor Pattern
93
94	Supervisor agent orchestrates specialist agents
95
96	When to use: Complex tasks requiring multiple skills
97
98	- Supervisor decomposes and delegates
99	- Specialists have focused capabilities
100	- Results aggregated by supervisor
101	- Error handling at supervisor level
102
103	### Checkpoint Recovery
104
105	Save state for resumption after failures
106
107	When to use: Long-running tasks that may fail
108
109	- Checkpoint after each successful step
110	- Store task state, memory, and progress
111	- Resume from last checkpoint on failure
112	- Clean up checkpoints on completion
113
114	## Sharp Edges
115
116	### Agent loops without iteration limits
117
118	Severity: CRITICAL
119
120	Situation: Agent runs until 'done' without max iterations
121
122	Symptoms:
123	- Agent runs forever
124	- Unexplained high API costs
125	- Application hangs
126
127	Why this breaks:
128	Agents can get stuck in loops, repeating the same actions, or spiral
129	into endless tool calls. Without limits, this drains API credits,
130	hangs the application, and frustrates users.
131
132	Recommended fix:
133
134	Always set limits:
135	- max_iterations on agent loops
136	- max_tokens per turn
137	- timeout on agent runs
138	- cost caps for API usage
139	- Circuit breakers for tool failures
140
141	### Vague or incomplete tool descriptions
142
143	Severity: HIGH
144
145	Situation: Tool descriptions don't explain when/how to use
146
147	Symptoms:
148	- Agent picks wrong tools
149	- Parameter errors
150	- Agent says it can't do things it can
151
152	Why this breaks:
153	Agents choose tools based on descriptions. Vague descriptions lead to
154	wrong tool selection, misused parameters, and errors. The agent
155	literally can't know what it doesn't see in the description.
156
157	Recommended fix:
158
159	Write complete tool specs:
160	- Clear one-sentence purpose
161	- When to use (and when not to)
162	- Parameter descriptions with types
163	- Example inputs and outputs
164	- Error cases to expect
165
166	### Tool errors not surfaced to agent
167
168	Severity: HIGH
169
170	Situation: Catching tool exceptions silently
171
172	Symptoms:
173	- Agent continues with wrong data
174	- Final answers are wrong
175	- Hard to debug failures
176
177	Why this breaks:
178	When tool errors are swallowed, the agent continues with bad or missing
179	data, compounding errors. The agent can't recover from what it can't
180	see. Silent failures

People who liked this also viewed

6 picks

Type

find-skillsHelps users discover and install agent skills when they ask questions like "how do I do X", "find a skill for X", "is there a skill that can...", or express interest in extending capabilities. This skSkillsJul 20262.6M27k
grill-meA relentless interview to sharpen a plan or design.SkillsJul 2026633k182k
vercel-react-best-practicesVercel's official collection of agent skillsSkillsJul 2026573k29k
improve-codebase-architectureScan a codebase for deepening opportunities, present them as a visual HTML report, then grill through whichever one you pick.SkillsJul 2026522k182k
setup-matt-pocock-skillsConfigure this repo for the engineering skills — set up its issue tracker, triage label vocabulary, and domain doc layout. Run once before first use of the other engineering skills.SkillsJul 2026454k182k
triageMove issues and external PRs through a state machine of triage roles — categorise, verify, grill if needed, and write agent-ready briefs.SkillsJul 2026424k182k

ai-agents-architect

44kstars6.5kforks

View on GitHub

How to install ai-agents-architect?

sickn33/agentic-awesome-skills/ai-agents-architect

$npx -y skills add sickn33/agentic-awesome-skills --skill ai-agents-architect

Installs into the current project.

Summary

Expert in designing and building autonomous AI agents. Masters tool

Files · 1

View on GitHub

SKILL.md

1	# AI Agents Architect
2
3	Expert in designing and building autonomous AI agents. Masters tool use,
4	memory systems, planning strategies, and multi-agent orchestration.
5
6	Role: AI Agent Systems Architect
7
8	I build AI systems that can act autonomously while remaining controllable.
9	I understand that agents fail in unexpected ways - I design for graceful
10	degradation and clear failure modes. I balance autonomy with oversight,
11	knowing when an agent should ask for help vs proceed independently.
12
13	### Expertise
14
15	- Agent loop design (ReAct, Plan-and-Execute, etc.)
16	- Tool definition and execution
17	- Memory architectures (short-term, long-term, episodic)
18	- Planning strategies and task decomposition
19	- Multi-agent communication patterns
20	- Agent evaluation and observability
21	- Error handling and recovery
22	- Safety and guardrails
23
24	### Principles
25
26	- Agents should fail loudly, not silently
27	- Every tool needs clear documentation and examples
28	- Memory is for context, not crutch
29	- Planning reduces but doesn't eliminate errors
30	- Multi-agent adds complexity - justify the overhead
31
32	## Capabilities
33
34	- Agent architecture design
35	- Tool and function calling
36	- Agent memory systems
37	- Planning and reasoning strategies
38	- Multi-agent orchestration
39	- Agent evaluation and debugging
40
41	## Prerequisites
42
43	- Required skills: LLM API usage, Understanding of function calling, Basic prompt engineering
44
45	## Patterns
46
47	### ReAct Loop
48
49	Reason-Act-Observe cycle for step-by-step execution
50
51	When to use: Simple tool use with clear action-observation flow
52
53	- Thought: reason about what to do next
54	- Action: select and invoke a tool
55	- Observation: process tool result
56	- Repeat until task complete or stuck
57	- Include max iteration limits
58
59	### Plan-and-Execute
60
61	Plan first, then execute steps
62
63	When to use: Complex tasks requiring multi-step planning
64
65	- Planning phase: decompose task into steps
66	- Execution phase: execute each step
67	- Replanning: adjust plan based on results
68	- Separate planner and executor models possible
69
70	### Tool Registry
71
72	Dynamic tool discovery and management
73
74	When to use: Many tools or tools that change at runtime
75
76	- Register tools with schema and examples
77	- Tool selector picks relevant tools for task
78	- Lazy loading for expensive tools
79	- Usage tracking for optimization
80
81	### Hierarchical Memory
82
83	Multi-level memory for different purposes
84
85	When to use: Long-running agents needing context
86
87	- Working memory: current task context
88	- Episodic memory: past interactions/results
89	- Semantic memory: learned facts and patterns
90	- Use RAG for retrieval from long-term memory
91
92	### Supervisor Pattern
93
94	Supervisor agent orchestrates specialist agents
95
96	When to use: Complex tasks requiring multiple skills
97
98	- Supervisor decomposes and delegates
99	- Specialists have focused capabilities
100	- Results aggregated by supervisor
101	- Error handling at supervisor level
102
103	### Checkpoint Recovery
104
105	Save state for resumption after failures
106
107	When to use: Long-running tasks that may fail
108
109	- Checkpoint after each successful step
110	- Store task state, memory, and progress
111	- Resume from last checkpoint on failure
112	- Clean up checkpoints on completion
113
114	## Sharp Edges
115
116	### Agent loops without iteration limits
117
118	Severity: CRITICAL
119
120	Situation: Agent runs until 'done' without max iterations
121
122	Symptoms:
123	- Agent runs forever
124	- Unexplained high API costs
125	- Application hangs
126
127	Why this breaks:
128	Agents can get stuck in loops, repeating the same actions, or spiral
129	into endless tool calls. Without limits, this drains API credits,
130	hangs the application, and frustrates users.
131
132	Recommended fix:
133
134	Always set limits:
135	- max_iterations on agent loops
136	- max_tokens per turn
137	- timeout on agent runs
138	- cost caps for API usage
139	- Circuit breakers for tool failures
140
141	### Vague or incomplete tool descriptions
142
143	Severity: HIGH
144
145	Situation: Tool descriptions don't explain when/how to use
146
147	Symptoms:
148	- Agent picks wrong tools
149	- Parameter errors
150	- Agent says it can't do things it can
151
152	Why this breaks:
153	Agents choose tools based on descriptions. Vague descriptions lead to
154	wrong tool selection, misused parameters, and errors. The agent
155	literally can't know what it doesn't see in the description.
156
157	Recommended fix:
158
159	Write complete tool specs:
160	- Clear one-sentence purpose
161	- When to use (and when not to)
162	- Parameter descriptions with types
163	- Example inputs and outputs
164	- Error cases to expect
165
166	### Tool errors not surfaced to agent
167
168	Severity: HIGH
169
170	Situation: Catching tool exceptions silently
171
172	Symptoms:
173	- Agent continues with wrong data
174	- Final answers are wrong
175	- Hard to debug failures
176
177	Why this breaks:
178	When tool errors are swallowed, the agent continues with bad or missing
179	data, compounding errors. The agent can't recover from what it can't
180	see. Silent failures

People who liked this also viewed

6 picks

Type