feat: add AWS, GCP, and Azure skill files - #288
Closed
beng360 wants to merge 5 commits into
Closed
Conversation
The agent can now load_skill('aws'), load_skill('gcp'), or
load_skill('azure') to get provider-specific CLI commands, Terraform
examples, error recovery steps, and critical rules — instead of
relying solely on the generic cloud_access.md core prompt.
Each skill follows the same structure as OVH/Scaleway:
- Discovery, compute, Kubernetes, storage, database, networking commands
- Terraform resource examples (provider block auto-generated)
- Multi-account support (AWS), resource group rules (Azure), project setup (GCP)
- Error recovery playbook per provider
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Contributor
📝 WalkthroughWalkthroughRestricts cloud command patterns to provider-specific CLI prefixes, mandates invoking load_skill('') on cloud_exec failures, and adds/rewrites provider skill documents for AWS, Azure, GCP, OVH, and Scaleway to be CLI-first with expanded CLI references and RCA workflows. Changes
Sequence Diagram(s)sequenceDiagram
participant Agent as Agent/System
participant CloudExec as cloud_exec()
participant SkillRegistry as load_skill()
participant Provider as Provider CLI
Agent->>CloudExec: Execute cloud command<br/>(provider-specific CLI)
CloudExec->>Provider: Run command
Provider-->>CloudExec: Failure (syntax/flag/error)
CloudExec-->>Agent: Failure response
Agent->>SkillRegistry: load_skill('<provider>')
SkillRegistry-->>Agent: Provider skill (CLI rules/docs)
Agent->>CloudExec: Retry command (apply provider skill guidance)
CloudExec->>Provider: Run corrected command
Provider-->>CloudExec: Success
CloudExec-->>Agent: Result
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~25 minutes Possibly related PRs
Suggested reviewers
🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
Rewrite AWS, GCP, and Azure SKILL.md files from surface-level cheat sheets into authoritative references the agent actually needs: - Full CLI reference per service (EC2/EKS/RDS/S3/Lambda/CloudWatch, Compute/GKE/SQL/Storage/Run/Logging, VMs/AKS/SQL/Storage/AppService) - CloudWatch Insights query patterns, Cloud Logging filter syntax, KQL Log Analytics queries — real investigation tooling - Complete Terraform examples sourced from Context7 provider docs (aws_eks_cluster with IAM, google_container_cluster with node pools, azurerm_kubernetes_cluster with identity) - RCA investigation workflows per provider (10+ step playbooks) - Error recovery with Context7 library IDs for each provider - Region mapping per provider - Provider-specific gotchas (AWS multi-account, Azure resource groups, GCP API enablement) Trim cloud_access.md from 73 lines to 10 — just the cloud_exec signature and "load the skill for details". The skills are now the authoritative source; the core prompt just explains the tool exists. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…ered skill loading - Rewrite OVH SKILL.md (125→305 lines): managed databases, block storage, load balancers, container registry, DNS, private networks, full RCA workflow - Rewrite Scaleway SKILL.md (132→364 lines): security groups, VPC, load balancers, serverless, secret manager, DNS, full RCA workflow - Remove Terraform/iac_tool from all 5 cloud provider skills (unused) - cloud_access.md: trigger load_skill on failure instead of suggesting it upfront - error_handling.md: add CLOUD COMMAND FAILURES rule to load skill before retry Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
beng360
commented
Apr 20, 2026
beng360
left a comment
beng360
left a comment
Contributor
Author
There was a problem hiding this comment.
Self-review: 5 inline findings — 1 minor formatting bug to fix, 4 observations for awareness.
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
Contributor
There was a problem hiding this comment.
Actionable comments posted: 2
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In `@server/chat/backend/agent/skills/integrations/ovh/SKILL.md`:
- Around line 115-135: The docs show kubectl invocations using terminal_exec
which violates the tool contract; replace each terminal_exec(...) example with
cloud_exec('ovh', '<kubectl command>') calls so the provider command path is
used (e.g. change terminal_exec('kubectl --kubeconfig=/tmp/kubeconfig.yaml get
pods -n <NS> -o wide') to cloud_exec('ovh', 'kubectl
--kubeconfig=/tmp/kubeconfig.yaml get pods -n <NS> -o wide') and do the same for
describe, logs, get events, top, get hpa/deployments/pvc/svc/ingress and rollout
history); keep the kubeconfig generation step using cloud_exec('ovh', 'cloud
kube kubeconfig generate <CLUSTER_ID> ...', output_file='/tmp/kubeconfig.yaml')
intact.
In `@server/chat/backend/agent/skills/integrations/scaleway/SKILL.md`:
- Around line 357-360: The Context7 lookup in the CLI troubleshooting section
uses the Terraform provider ID string '/scaleway/terraform-provider-scaleway'
via the mcp_context7_get_library_docs call, which is inappropriate for CLI
failures; update the CLI failure guidance by either removing that Context7
reference or replacing it with a Scaleway CLI-specific Context7 ID (if
available), and ensure the section instead prioritizes the existing `--help`
fallback advice mentioned earlier; search for the literal
`mcp_context7_get_library_docs` call and the string
'/scaleway/terraform-provider-scaleway' in SKILL.md and modify accordingly.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Repository UI
Review profile: CHILL
Plan: Pro
Run ID: 1be6ec83-ca05-428f-9159-f6f5c085e0f6
📒 Files selected for processing (7)
server/chat/backend/agent/skills/core/cloud_access.mdserver/chat/backend/agent/skills/core/error_handling.mdserver/chat/backend/agent/skills/integrations/aws/SKILL.mdserver/chat/backend/agent/skills/integrations/azure/SKILL.mdserver/chat/backend/agent/skills/integrations/gcp/SKILL.mdserver/chat/backend/agent/skills/integrations/ovh/SKILL.mdserver/chat/backend/agent/skills/integrations/scaleway/SKILL.md
…7, Azure formatting
- OVH SKILL.md: change all 14 kubectl `terminal_exec` calls to `cloud_exec('ovh', 'kubectl ...')`
- Scaleway SKILL.md: fix Context7 library ID from terraform-provider to scaleway-cli
- Azure SKILL.md: add missing blank line between Subscription Setup and CLI Reference sections
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Contributor
There was a problem hiding this comment.
🧹 Nitpick comments (3)
server/chat/backend/agent/skills/integrations/scaleway/SKILL.md (1)
348-360: Addload_skill('scaleway')as first retry action forcloud_execfailures.The recovery flow currently jumps to syntax/help and Context7. Add the required provider skill reload step first to keep runtime behavior consistent with core error handling.
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@server/chat/backend/agent/skills/integrations/scaleway/SKILL.md` around lines 348 - 360, Add a first retry step that calls load_skill('scaleway') whenever cloud_exec failures occur: update the Error Recovery flow in SKILL.md so that immediately after general checks (resource/quotas/permissions/etc.) and before the "CLI syntax" step and the Context7 lookup, you insert a retry action to reload the provider skill by invoking load_skill('scaleway') (referencing the cloud_exec failure handling and the existing Context7 lookup mcp_context7_get_library_docs entry) so runtime behavior matches core error handling.server/chat/backend/agent/skills/integrations/azure/SKILL.md (1)
233-244: Add explicitload_skill('azure')retry step in Error Recovery.The section should explicitly state that when a
cloud_execcommand fails, retry only afterload_skill('azure'), to match the core failure-handling contract and reduce inconsistent recovery behavior.Suggested doc patch
## Error Recovery +0. **Any cloud_exec failure** → Call `load_skill('azure')` before retrying command syntax/flags. 1. **Provider not registered** → `cloud_exec('azure', 'provider register --namespace Microsoft.ContainerService')` — common: Microsoft.ContainerService, Microsoft.Sql, Microsoft.Storage, Microsoft.Web, Microsoft.Network 2. **Resource group missing** → Create one: `group create --name <RG> --location <LOC>`🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@server/chat/backend/agent/skills/integrations/azure/SKILL.md` around lines 233 - 244, Update the Error Recovery section to require an explicit retry only after calling load_skill('azure') when any cloud_exec('azure', ...) command fails: mention load_skill('azure') as the first recovery step before retrying the failing cloud_exec command, and update the CLI-errors paragraph (including the "For CLI errors" Context7 example with mcp_context7_get_library_docs) to note that retries must follow load_skill('azure') to satisfy the core failure-handling contract; reference cloud_exec and load_skill('azure') by name so readers know where to apply the retry.server/chat/backend/agent/skills/integrations/ovh/SKILL.md (1)
284-296: Include mandatoryload_skill('ovh')before retrying failed cloud commands.Line 291 and the Context7 fallback are useful, but this section should first require
load_skill('ovh')oncloud_execfailures to align with core error-handling behavior.🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@server/chat/backend/agent/skills/integrations/ovh/SKILL.md` around lines 284 - 296, Add a mandatory step to call load_skill('ovh') before retrying any failed cloud_exec calls: update the "Error Recovery" list and the "Context7 lookup on failure" paragraph to instruct operators to run load_skill('ovh') when a cloud_exec('<COMMAND>') returns an error, then proceed to the existing fallback (e.g., mcp_context7_get_library_docs(context7CompatibleLibraryID='/ovh/ovhcloud-cli', topic='cloud instance create')) and CLI --help checks; reference cloud_exec and load_skill('ovh') explicitly so readers know to load the ovh skill prior to retrying or invoking Context7.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Nitpick comments:
In `@server/chat/backend/agent/skills/integrations/azure/SKILL.md`:
- Around line 233-244: Update the Error Recovery section to require an explicit
retry only after calling load_skill('azure') when any cloud_exec('azure', ...)
command fails: mention load_skill('azure') as the first recovery step before
retrying the failing cloud_exec command, and update the CLI-errors paragraph
(including the "For CLI errors" Context7 example with
mcp_context7_get_library_docs) to note that retries must follow
load_skill('azure') to satisfy the core failure-handling contract; reference
cloud_exec and load_skill('azure') by name so readers know where to apply the
retry.
In `@server/chat/backend/agent/skills/integrations/ovh/SKILL.md`:
- Around line 284-296: Add a mandatory step to call load_skill('ovh') before
retrying any failed cloud_exec calls: update the "Error Recovery" list and the
"Context7 lookup on failure" paragraph to instruct operators to run
load_skill('ovh') when a cloud_exec('<COMMAND>') returns an error, then proceed
to the existing fallback (e.g.,
mcp_context7_get_library_docs(context7CompatibleLibraryID='/ovh/ovhcloud-cli',
topic='cloud instance create')) and CLI --help checks; reference cloud_exec and
load_skill('ovh') explicitly so readers know to load the ovh skill prior to
retrying or invoking Context7.
In `@server/chat/backend/agent/skills/integrations/scaleway/SKILL.md`:
- Around line 348-360: Add a first retry step that calls load_skill('scaleway')
whenever cloud_exec failures occur: update the Error Recovery flow in SKILL.md
so that immediately after general checks (resource/quotas/permissions/etc.) and
before the "CLI syntax" step and the Context7 lookup, you insert a retry action
to reload the provider skill by invoking load_skill('scaleway') (referencing the
cloud_exec failure handling and the existing Context7 lookup
mcp_context7_get_library_docs entry) so runtime behavior matches core error
handling.
ℹ️ Review info
⚙️ Run configuration
Configuration used: Repository UI
Review profile: CHILL
Plan: Pro
Run ID: 43a4f420-8c6a-4fd6-abc0-3b76781d0ef2
📒 Files selected for processing (3)
server/chat/backend/agent/skills/integrations/azure/SKILL.mdserver/chat/backend/agent/skills/integrations/ovh/SKILL.mdserver/chat/backend/agent/skills/integrations/scaleway/SKILL.md
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
load_skill('aws')/load_skill('gcp')/load_skill('azure')on demandcloud_access.md— the agent had no provider-specific skill to load when users asked about AWS/GCP/Azure, andload_skill('aws')returned "No skill 'aws'"What each skill covers
Test plan
load_skill('aws')returns skill content for connected AWS usersload_skill('gcp')returns skill content for connected GCP usersload_skill('azure')returns skill content for connected Azure users🤖 Generated with Claude Code
Summary by CodeRabbit
New Features
Documentation