Back to skill

Security audit

huawei-cloud-cloudrobo-dataset

Security checks for vulnerabilities and agentic risk

Overview

The skill fits its cloud-data purpose, but credential exposure, overly broad cleanup, and an unsafe remote web-server option warrant review.

Request fixes for secret-bearing diagnostics, cleanup outside the current run, conflicting Mode B push instructions, and remote recording access. Use an isolated build host, narrowly scoped cloud credentials, verified TLS, and the loopback web launcher. Avoid sharing complete login commands and review test-runner substitutions before execution. I independently investigated all six partially covered files, including installation, cleanup, web data flows, and test commands. Missing-reference examples were runtime filenames or field names; the flagged documentation showed no encoded instruction payload in the checks performed. External dependencies and runtime behavior remain unverified.

Vulnerability Patterns
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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
Findings (1)

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/bag2manifest/bag2manifest/webapp.py
Finding

Remote clients can profile server-side recordings outside their uploaded files

Content
View full analysis

Impact

The visitor can obtain topic metadata, numeric samples and camera thumbnails from readable server-side ROS recordings through the analysis results. Successful export also includes profile.json. This finding does not establish unrestricted raw-file reading or credential theft.

Vulnerability Patterns
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • Tool MisuseTool Parameter Abuse, Chaining Abuse, Unsafe Defaults
  • Behavioral ASTexec() Call, eval() Call, Dynamic Import
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (123)

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 73)May include surrounding context.

md
See `references/cli-installation-guide.md` for CLI installation, AK/SK authentication, and

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 649)May include surrounding context.

md
- [CLI Installation Guide](references/cli-installation-guide.md) — cloudrobo CLI installation and configuration

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 145)May include surrounding context.

md
erator failures** → fetch `job-std-output.log` and apply the P0-P3 patterns in `references/custom-operator/custom-operator-workflow.md` §Failure Analysis Guide.

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 592)May include surrounding context.

md
provide a full `docker login` command, see `references/custom-operator/custom-operator-workflow.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 624)May include surrounding context.

md
- [Custom Operator Workflow](references/custom-operator/custom-operator-workflow.md) — Phase 0-5 steps, phase boundaries, failure-pattern matrix (P0-P3), failur

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 657)May include surrounding context.

md
- [Custom Operator Workflow](references/custom-operator/custom-operator-workflow.md) — five-phase onboarding + phase boundaries + failure patterns (incl. 5d HTT

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 755)May include surrounding context.

md
only this run's artifacts (never customer images/containers/data). Full rules: `references/custom-operator/custom-operator-workflow.md` + `references/custom-ope

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 · SKILL.md (reported line 243)May include surrounding context.

text

- **SDK:** `client.delete_tasks([task_id_1, task_id_2])` (accepts a list; CLI deletes one at a time)
- **API:** `DELETE /v1/data-eng/proc-tasks?ids=id1,id2`

#### Copy a task

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 · SKILL.md (reported line 422)May include surrounding context.

text

- **SDK:** `client.delete_eval_task(task_id)`
- **API:** `DELETE /v1/data-eng/eval-tasks/{task_id}`

**Note:** eval-tasks deletion is single-task granularity (`delete-task`, not batch). eval-tasks do
not support restart — to re-run, use `copy-task`.

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 625)May include surrounding context.

md
- [Custom Operator Reference](references/custom-operator/custom-operator-reference.md) — operator_config catalog, Mode A/B, dialogue control layer, report templ

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 639)May include surrounding context.

md
对话逻辑唯一权威定义在 `references/custom-operator/custom-operator-reference.md`(System

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 658)May include surrounding context.

md
- [Custom Operator Reference](references/custom-operator/custom-operator-reference.md) — operator_config catalog + Mode A/B + dialogue control layer

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 755)May include surrounding context.

md
/data). Full rules: `references/custom-operator/custom-operator-workflow.md` + `references/custom-operator/custom-operator-reference.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 653)May include surrounding context.

md
- [Acceptance Criteria](references/acceptance-criteria.md) — Acceptance criteria

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 726)May include surrounding context.

md
See `templates/test-vars.json` for the full test case list covering proc-tasks, eval-tasks,

Chaining Abuse

High
Category
Tool Misuse
Confidence
98% confidence
Finding

The pipe directly connects remote content to a privileged shell, allowing the downloaded response to determine root-level actions. This is the same unsafe installation operation flagged by the script-fetching and privileged-execution findings, not evidence of a separate malicious chain.

Content

Scanner excerpt · references/cli-installation-guide.md (reported line 53)May include surrounding context.

bash
# Ubuntu/Debian
curl -fsSL https://get.docker.com | sudo bash
sudo systemctl start docker
sudo usermod -aG docker $USER  # Avoid sudo for docker commands

L55 instructs adding the user to the Docker group without warning that this typically grants root-equivalent privileges.

High
Category
Output Handling
Confidence
98% confidence
Finding

Adding the user to the Docker group typically grants root-equivalent control through the Docker daemon. Presenting this as a convenience without explaining the privilege change can expose the host to compromise by processes running as that user, including operator build tooling.

Content

No source excerpt is available for this finding.

Chaining Abuse

High
Category
Tool Misuse
Confidence
75% confidence
Finding

Tool calls are chained to bypass individual safety checks or escalate capabilities beyond what any single tool call would allow.

Content

Scanner excerpt · references/custom-operator/custom-operator-reference.md (reported line 488)May include surrounding context.

fail. Fix (once, on the preset image):

dockerfile
RUN chmod o+rx /opt/cloud  \
 && chmod -R o+rX /opt/cloud/3rdComponent \
 && chmod o+x /opt/cloud/3rdComponent/python/bin/python3.12

Verify as the most restrictive user before pushing:

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
90% 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/custom-operator/custom-operator-reference.md (reported line 677)May include surrounding context.

text
- **pip with `--no-cache-dir`** and clean pip cache:
  ```dockerfile
  RUN python3 -m pip install --no-cache-dir -r requirements.txt && rm -rf ~/.cache/pip
  • Multi-stage build when the operator needs build-time-only tools (gcc, headers): build wheels in a builder stage, copy only wheels/site-packages into the final

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
90% 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 · templates/Dockerfile (reported line 57)May include surrounding context.

dockerfile
# (see references/custom-operator/custom-operator-reference.md §Verified production lessons, lessons 6–7).
COPY requirements.txt .
RUN python3 -m pip install --no-cache-dir -r requirements.txt 2>/dev/null \
    && rm -rf ~/.cache/pip \
    || echo "pip not available in base image, continuing with stdlib only"

# ── Layer 3: User source code ────────────────────────────────────────

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
90% 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/custom-operator/custom-operator-reference.md (reported line 677)May include surrounding context.

text
- **pip with `--no-cache-dir`** and clean pip cache:
  ```dockerfile
  RUN python3 -m pip install --no-cache-dir -r requirements.txt && rm -rf ~/.cache/pip
  • Multi-stage build when the operator needs build-time-only tools (gcc, headers): build wheels in a builder stage, copy only wheels/site-packages into the final

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
90% 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 · templates/Dockerfile (reported line 57)May include surrounding context.

dockerfile
# (see references/custom-operator/custom-operator-reference.md §Verified production lessons, lessons 6–7).
COPY requirements.txt .
RUN python3 -m pip install --no-cache-dir -r requirements.txt 2>/dev/null \
    && rm -rf ~/.cache/pip \
    || echo "pip not available in base image, continuing with stdlib only"

# ── Layer 3: User source code ────────────────────────────────────────

Chaining Abuse

High
Category
Tool Misuse
Confidence
75% confidence
Finding

Tool calls are chained to bypass individual safety checks or escalate capabilities beyond what any single tool call would allow.

Content

Scanner excerpt · references/custom-operator/custom-operator-reference.md (reported line 677)May include surrounding context.

text
- **pip with `--no-cache-dir`** and clean pip cache:
  ```dockerfile
  RUN python3 -m pip install --no-cache-dir -r requirements.txt && rm -rf ~/.cache/pip
  • Multi-stage build when the operator needs build-time-only tools (gcc, headers): build wheels in a builder stage, copy only wheels/site-packages into the final

Chaining Abuse

High
Category
Tool Misuse
Confidence
75% confidence
Finding

Tool calls are chained to bypass individual safety checks or escalate capabilities beyond what any single tool call would allow.

Content

Scanner excerpt · templates/Dockerfile (reported line 33)May include surrounding context.

dockerfile
#   dnf/yum-based (HCE/EulerOS):
# RUN dnf install -y gcc gcc-c++ make \
#     && dnf clean all \
#     && rm -rf /var/cache/dnf
# NOTE (HCE): to build C-extension deps (e.g. evdev via pynput) you ALSO need
#   binutils-extra (HCE binutils ships WITHOUT ld), kernel-headers, python3-devel:
# RUN dnf install -y binutils binutils-extra gcc make kernel-headers python3-devel \

Chaining Abuse

High
Category
Tool Misuse
Confidence
75% confidence
Finding

Tool calls are chained to bypass individual safety checks or escalate capabilities beyond what any single tool call would allow.

Content

Scanner excerpt · templates/Dockerfile (reported line 37)May include surrounding context.

dockerfile
# NOTE (HCE): to build C-extension deps (e.g. evdev via pynput) you ALSO need
#   binutils-extra (HCE binutils ships WITHOUT ld), kernel-headers, python3-devel:
# RUN dnf install -y binutils binutils-extra gcc make kernel-headers python3-devel \
#     && dnf clean all && rm -rf /var/cache/dnf
#   and remove them afterwards to slim the image (see references/custom-operator/custom-operator-reference.md
#   §Verified production lessons, lesson 5).
# NOTE (network): `docker build` RUN steps do NOT inherit the daemon proxy — pass