agents-harden

Name: agents-harden
Author: aws

575installs190forks

How to install agents-harden?

aws/agent-toolkit-for-aws/agents-harden

$npx -y skills add aws/agent-toolkit-for-aws --skill agents-harden

Installs into the current project.

Summary

Use when preparing your agent for production — IAM scoping, inbound auth (JWT, SigV4), secrets management, cold start optimization, session lifecycle, rate limiting, input validation, and quota guidance. Triggers on: "production checklist", "harden agent", "production ready", "se

Files · 1

View on GitHub

SKILL.md

1	# harden
2
3	Prepare your AgentCore agent for production — security, reliability, and performance.
4
5	## When to use
6
7	- You're about to take an agent to production
8	- You want a checklist of what to review before launch
9	- You want to restrict who can call your agent
10	- You want to scope down IAM permissions from the defaults
11	- You're hitting throttling or quota errors (loads [`references/limits.md`](references/limits.md))
12	- You need to tune session lifecycle for your workload
13	- You're running long-running background work in your agent
14
15	## Input
16
17	No arguments required. The skill reads your project config and produces a checklist with specific findings for your project.
18
19	## Process
20
21	### Step 0: Verify CLI version
22
23	Run `agentcore --version`. This skill requires v0.9.0 or later. If the version is older, tell the developer to run `agentcore update` before proceeding.
24
25	### Step 1: Read the project
26
27	Read `agentcore/agentcore.json` to understand:
28
29	- What resources are configured (memory, gateway, credentials, evaluators)
30	- What framework is being used
31	- What network mode is configured (PUBLIC or VPC)
32
33	### Step 2: Run through the checklist
34
35	Work through each category and report findings specific to the project.
36
37	---
38
39	## IAM: Scope down permissions
40
41	The auto-created execution role has broad Bedrock access (`arn:aws:bedrock:::foundation-model/`). For production, scope it to the specific models your agent uses.
42
43	Check the current execution role:
44
45	```bash
46	agentcore status --json \| jq -r '.runtimes[0].executionRoleArn'
47	```
48
49	Recommended production Bedrock policy:
50
51	```json
52	{
53	"Effect": "Allow",
54	"Action": [
55	"bedrock:InvokeModel",
56	"bedrock:InvokeModelWithResponseStream"
57	],
58	"Resource": [
59	"arn:aws:bedrock:<REGION>::foundation-model/anthropic.claude-sonnet-4-5-20250929-v1:0"
60	]
61	}
62	```
63
64	Replace the resource ARN with the specific model(s) your agent uses.
65
66	ECR access: Scope to your specific repository:
67
68	```json
69	{
70	"Effect": "Allow",
71	"Action": ["ecr:BatchGetImage", "ecr:GetDownloadUrlForLayer"],
72	"Resource": "arn:aws:ecr:<REGION>:<YOUR_ACCOUNT_ID>:repository/bedrock-agentcore-<AGENT_NAME>-*"
73	}
74	```
75
76	Trust policy: Verify the execution role's trust policy is scoped to your account:
77
78	```json
79	{
80	"Principal": {"Service": "bedrock-agentcore.amazonaws.com"},
81	"Action": "sts:AssumeRole",
82	"Condition": {
83	"StringEquals": {"aws:SourceAccount": "<YOUR_ACCOUNT_ID>"},
84	"ArnLike": {"aws:SourceArn": "arn:aws:bedrock-agentcore:<REGION>:<YOUR_ACCOUNT_ID>:*"}
85	}
86	}
87	```
88
89	Runtime resource-based policies (API-only): For fine-grained control over which principals can invoke your runtime — beyond what IAM roles and JWT auth provide — use `PutAgentRuntimeResourcePolicy` via boto3. This is not exposed in the CLI or `agentcore.json`. Use the `awsknowledge` MCP server if available to look up the current API shape.
90
91	---
92
93	## Shell Access: Scope `InvokeAgentRuntimeCommand` separately
94
95	If your project uses `InvokeAgentRuntimeCommand` (see [`agents-build/references/integrate.md`](../agents-build/references/integrate.md)), audit its IAM permissions separately from `InvokeAgentRuntime`. The two actions have different blast radii: `InvokeAgentRuntimeCommand` is arbitrary shell execution inside a live microVM with the runtime's full execution role — callers can read/write the filesystem, reach any network resource the agent can reach, and access the execution role's credentials.
96
97	Check which principals have the permission:
98
99	```bash
100	# List customer-managed policies in your account, then inspect each for InvokeAgentRuntimeCommand
101	aws iam list-policies --scope Local \
102	--query 'Policies[*].[PolicyName, Arn, DefaultVersionId]' \
103	--output table
104	# Then for each policy of interest:
105	aws iam get-policy-version \
106	--policy-arn <POLICY_ARN> \
107	--version-id <VERSION_ID> \
108	--query 'PolicyVersion.Document'
109	```
110
111	Alternatively, use the

People who liked this also viewed

6 picks

Type

agents-harden

575installs190forks

View on GitHub

How to install agents-harden?

aws/agent-toolkit-for-aws/agents-harden

$npx -y skills add aws/agent-toolkit-for-aws --skill agents-harden

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# harden
2
3	Prepare your AgentCore agent for production — security, reliability, and performance.
4
5	## When to use
6
7	- You're about to take an agent to production
8	- You want a checklist of what to review before launch
9	- You want to restrict who can call your agent
10	- You want to scope down IAM permissions from the defaults
11	- You're hitting throttling or quota errors (loads [`references/limits.md`](references/limits.md))
12	- You need to tune session lifecycle for your workload
13	- You're running long-running background work in your agent
14
15	## Input
16
17	No arguments required. The skill reads your project config and produces a checklist with specific findings for your project.
18
19	## Process
20
21	### Step 0: Verify CLI version
22
23	Run `agentcore --version`. This skill requires v0.9.0 or later. If the version is older, tell the developer to run `agentcore update` before proceeding.
24
25	### Step 1: Read the project
26
27	Read `agentcore/agentcore.json` to understand:
28
29	- What resources are configured (memory, gateway, credentials, evaluators)
30	- What framework is being used
31	- What network mode is configured (PUBLIC or VPC)
32
33	### Step 2: Run through the checklist
34
35	Work through each category and report findings specific to the project.
36
37	---
38
39	## IAM: Scope down permissions
40
41	The auto-created execution role has broad Bedrock access (`arn:aws:bedrock:::foundation-model/`). For production, scope it to the specific models your agent uses.
42
43	Check the current execution role:
44
45	```bash
46	agentcore status --json \| jq -r '.runtimes[0].executionRoleArn'
47	```
48
49	Recommended production Bedrock policy:
50
51	```json
52	{
53	"Effect": "Allow",
54	"Action": [
55	"bedrock:InvokeModel",
56	"bedrock:InvokeModelWithResponseStream"
57	],
58	"Resource": [
59	"arn:aws:bedrock:<REGION>::foundation-model/anthropic.claude-sonnet-4-5-20250929-v1:0"
60	]
61	}
62	```
63
64	Replace the resource ARN with the specific model(s) your agent uses.
65
66	ECR access: Scope to your specific repository:
67
68	```json
69	{
70	"Effect": "Allow",
71	"Action": ["ecr:BatchGetImage", "ecr:GetDownloadUrlForLayer"],
72	"Resource": "arn:aws:ecr:<REGION>:<YOUR_ACCOUNT_ID>:repository/bedrock-agentcore-<AGENT_NAME>-*"
73	}
74	```
75
76	Trust policy: Verify the execution role's trust policy is scoped to your account:
77
78	```json
79	{
80	"Principal": {"Service": "bedrock-agentcore.amazonaws.com"},
81	"Action": "sts:AssumeRole",
82	"Condition": {
83	"StringEquals": {"aws:SourceAccount": "<YOUR_ACCOUNT_ID>"},
84	"ArnLike": {"aws:SourceArn": "arn:aws:bedrock-agentcore:<REGION>:<YOUR_ACCOUNT_ID>:*"}
85	}
86	}
87	```
88
89	Runtime resource-based policies (API-only): For fine-grained control over which principals can invoke your runtime — beyond what IAM roles and JWT auth provide — use `PutAgentRuntimeResourcePolicy` via boto3. This is not exposed in the CLI or `agentcore.json`. Use the `awsknowledge` MCP server if available to look up the current API shape.
90
91	---
92
93	## Shell Access: Scope `InvokeAgentRuntimeCommand` separately
94
95	If your project uses `InvokeAgentRuntimeCommand` (see [`agents-build/references/integrate.md`](../agents-build/references/integrate.md)), audit its IAM permissions separately from `InvokeAgentRuntime`. The two actions have different blast radii: `InvokeAgentRuntimeCommand` is arbitrary shell execution inside a live microVM with the runtime's full execution role — callers can read/write the filesystem, reach any network resource the agent can reach, and access the execution role's credentials.
96
97	Check which principals have the permission:
98
99	```bash
100	# List customer-managed policies in your account, then inspect each for InvokeAgentRuntimeCommand
101	aws iam list-policies --scope Local \
102	--query 'Policies[*].[PolicyName, Arn, DefaultVersionId]' \
103	--output table
104	# Then for each policy of interest:
105	aws iam get-policy-version \
106	--policy-arn <POLICY_ARN> \
107	--version-id <VERSION_ID> \
108	--query 'PolicyVersion.Document'
109	```
110
111	Alternatively, use the

People who liked this also viewed

6 picks

Type