Back to skill

Security audit

huawei-cloud-cloudrobo-infer

Security checks for vulnerabilities and agentic risk

Overview

ClawScan returns a benign verdict with medium confidence: most reviewed behavior was judged consistent with the skill’s stated purpose, though one referenced guide was only partly inspected and a reported asset-download behavior was not fully explained.

The verdict is benign, but benign does not mean risk-free, and scan coverage was incomplete. Before installing, consider that the scanners could not fully inspect the installation guide and did not establish how the reported asset-file retrieval supports the stated deployment purpose.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • Tool MisuseTool Parameter Abuse, Chaining Abuse, Unsafe Defaults
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (15)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
87% confidence
Finding

The code accesses /v1/assets/{asset_id}/versions/{version_id}/download-url and fetches the returned file URL. It performs no inference-service operation and establishes no connection between the downloaded asset and model deployment or service configuration. Although asset retrieval could support deployment, that supporting role is not demonstrated in the supplied chunk.

Content

No source excerpt is available for this finding.

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 71)May include surrounding context.

md
- CLI installed, AK/SK configured, workspace resolved — `references/cli-installation-guide.md`.

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 94)May include surrounding context.

md
echo "未检测到 cloudrobo 命令:请先安装聚合客户端 CLI(pip install hw-cloudrobo-client),详见 references/cli-installation-guide.md。" >&2; exit 2; }

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 345)May include surrounding context.

md
- [CLI Installation Guide](references/cli-installation-guide.md) — installation, AK/SK, region, workspace

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
80% confidence
Finding

Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).

Content

Scanner excerpt · references/iam-policies.md (reported line 55)May include surrounding context.

md
| Create service | Write access to `cloudrobo-service` (`POST /v1/infer-services`) |
| Show service detail | Read access to `cloudrobo-service` (`GET /v1/infer-services/{service_id}`) |
| Update service | Write access to `cloudrobo-service` (`PUT /v1/infer-services/{service_id}`) |
| Delete service | Delete access to `cloudrobo-service` (`DELETE /v1/infer-services/{service_id}`) |
| Start / stop service | Write access to `cloudrobo-service` (`POST /v1/infer-services/{service_id}/start`, `.../stop`) |
| List service logs | Read access to `cloudrobo-service` (`POST /v1/infer-services/{service_id}/logs`) |
| Query model asset (deploy source) | Read access to `cloudrobo-asset-manager` model assets |

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
80% confidence
Finding

Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).

Content

Scanner excerpt · references/service-config-catalog.md (reported line 174)May include surrounding context.

md
| list_infer_services | `list_infer_services(**params)` | `list` | `GET /v1/infer-services` |
| show_infer_service | `show_infer_service(service_id)` | `show` | `GET /v1/infer-services/{service_id}` |
| update_infer_service | `update_infer_service(service_id, req)` | `update` | `PUT /v1/infer-services/{service_id}` |
| delete_infer_service | `delete_infer_service(service_id)` | `delete` | `DELETE /v1/infer-services/{service_id}` |
| start_infer_service | `start_infer_service(service_id)` | `start` | `POST /v1/infer-services/{service_id}/start` |
| stop_infer_service | `stop_infer_service(service_id)` | `stop` | `POST /v1/infer-services/{service_id}/stop` |
| list_infer_service_logs | `list_infer_service_logs(service_id, req)` | `list-logs` | `POST /v1/infer-services/{service_id}/logs` |

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
85% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · SKILL.md (reported line 236)May include surrounding context.

md
2. **Feedback** — report the final status; on `FAILED` suggest `list-logs`.

> `wait-deploy` does NOT create the service — call `create` first, then `wait-deploy`. Do NOT
> call `start` after `create` (auto-deploys `CREATING → DEPLOYING → RUNNING`); `start` is only
> for restarting a `STOPPED` service.

### Service O&M / Lookup Workflow

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
85% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · SKILL.md (reported line 295)May include surrounding context.

md
5. A trained model is a **space asset** → Variant B of Step 7: run Step 5 auto-discovery and
   carry whatever was discovered.
6. `cloudrobo infer create ...` (Variant B) → `cloudrobo infer wait-deploy --service-id <sid>`
   (do NOT call `start` — the service auto-deploys).
7. Report the service `id` and status; the service is now consumable.

## Core Commands

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
85% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · SKILL.md (reported line 402)May include surrounding context.

md
- Always include `--stop-schedule-json '{"duration": N, "time_unit": "MINUTES"}'` in `create`;
  omit `--pool-id` unless deploying to a dedicated pool
- Use `wait-deploy` after `create` (5s interval, 600s default). Do NOT call `start` after
  `create` — the service auto-deploys; re-invoke `wait-deploy` if it returns while `CREATING`
- Use `--dry-run` on `create` to validate parameter assembly before submission
- Report health-check pickup as part of deployment verification
- Use `list-logs` with `--keywords` to isolate errors fast; remember **ms** timestamps

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
85% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · references/dataflow-diagram.md (reported line 63)May include surrounding context.

text

> **Note**: `wait-deploy` does NOT create the service. Call `create` first, then `wait-deploy`.
> Do NOT call `start` after `create` — the service auto-deploys (CREATING → DEPLOYING → RUNNING).
> `start` is only for restarting a `STOPPED` service.

## Log Diagnosis Dataflow

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
85% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · references/service-config-catalog.md (reported line 101)May include surrounding context.

md
- **ID validation**: `validate_safe_id(service_id, "service_id")` — empty/`..`/`/`/`\` rejected.

**Important**: `wait-deploy` does NOT create the service. Call `create` first, then `wait-deploy`.
Do NOT call `start` after `create` — the service auto-deploys (CREATING → DEPLOYING → RUNNING).
`start` is only for restarting a `STOPPED` service.

## Parameter Auto-Discovery Sources

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
85% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · references/verification-method.md (reported line 51)May include surrounding context.

md
cloudrobo infer create --name verify-infer --flavor "<flavor>" --model-json '{"model_id": "<asset-id>", "model_version_id": "<latest-version-id>"}' --workspace-id <ws-id> --pool-type <pool-type> --stop-schedule-json '{"duration": 60, "time_unit": "MINUTES"}' [--pool-id pool-<resource-id>]
# → Returns {"id": "..."} — service identifier field is `id` (some docs call it service_id); service auto-enters CREATING → DEPLOYING

# 7. Wait for deployment (do NOT call start — service auto-deploys)
cloudrobo infer wait-deploy --service-id <service-id> --timeout 600
# → Polls every 5s until status is no longer DEPLOYING (status != DEPLOYING)
#   (returns RUNNING/FAILED/...; if it returns while CREATING, re-invoke — it only waits on DEPLOYING)

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
85% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · references/verification-method.md (reported line 98)May include surrounding context.

Level 3: Wait-Deploy Test

bash
# First create the service (it auto-deploys), then wait for deployment
cloudrobo infer wait-deploy --service-id <service-id> --timeout 600
# → Polls every 5s, reports final status when status is no longer DEPLOYING (status != DEPLOYING)

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
85% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · references/service-config-catalog.md (reported line 226)May include surrounding context.

final = client.wait_deploy(sid, timeout=600) # RUNNING / FAILED / STOPPED ...

text

> Do NOT call `start_infer_service` after `create` — the service auto-deploys
> (`CREATING → DEPLOYING → RUNNING`); `start` only restarts a `STOPPED` service.

## Parameter Resolution & Confirmation

Preflight failures must be reported in Chinese without user language opt-in.

Low
Category
Output Handling
Confidence
99% confidence
Finding

Line 90 explicitly requires Chinese error messages on any preflight failure, and the following shell examples implement that requirement. The skill offers no language choice or justification for forcing Chinese regardless of the user's language.

Content

No source excerpt is available for this finding.