Huawei Cloud VoD Collector Skill: Report Bad Experiences
Captures bad experiences with Huawei Cloud tools, strips secrets and files them as GitCode issues for Huawei.
- Installs
- 17.5K
- on skills.sh
- Stars
- 59
- on GitHub
- Files
- 21
- besides SKILL.md
- Updated
- Oct 2026
- last repo commit
This is Huawei Cloud's own feedback skill, called Voice of Developer (VoD). When you hit an error, reject an answer, or say you want to report a problem, the agent writes a feedback file under .vod/feedbacks/ in your project, redacts secrets, and can file it as a GitCode issue for Huawei's teams.
Know three things before you install it. SKILL.md makes the agent wrap every script in a telemetry command, skill-quality-cli, that the skill installs for you. Filing an issue may need a GitCode sign-in through a separate open-source tool, AtomGit-GO, which the skill installs too. And the hook files that trigger capture automatically are provided for OpenClaw and Hermes only.
Good for
- Developers who hit errors with Huawei Cloud tools and want to report them with context
- OpenClaw or Hermes users who want errors and rejections captured automatically through the provided hooks
Install huawei-cloud-vod-collector
Claude Code
- 1
Copy this prompt.
Install the huawei-cloud-vod-collector skill in Claude Code. Run this command: npx skills add huaweicloud/huaweicloud-skills --skill huawei-cloud-vod-collector Then check that it loaded and tell me how to use it.
- 2
Paste it into Claude Code. It shows the change and asks before running it.
Requires Node.js.
Install the folder, then run the pip command with <SKILL_DIR> replaced by the folder that holds SKILL.md. The agent installs skill-quality-cli itself through scripts/ensure_cli.sh. The hook files are for OpenClaw and Hermes (references/hooks-setup.md); the repo has no plugin manifest, so there is no plugin route. There is no Claude app tab because the skill runs local scripts.
Built from the package data we hold. Not yet tested by us. Source.
What's in the skill
When it loads
Invoke this skill to capture poor experiences and distill them into high-value requirements (Voice of Developer). Use when user encounters any Huawei Cloud related issues, like user expresses dissatisfaction, encounters errors, or wants to report issues/suggestions.Triggers include: "体验差","反馈问题","反馈建议","这个有bug","拒绝了请求","报告问题","反馈体验","report a problem","report a suggestion","bug report","poor experience","voice of developer"
Sections in SKILL.md
- Overview
- Core Commands
- Parameter Confirmation
- References
- Prerequisites
- Python dependencies
- Workflow
- Phase 1: Capture
- Phase 2: Extract
- Phase 3: Deliver
- Behavioral Constraints
- Storage
- CLI Reference
- KooCLI region
- Token Configuration
Key facts
- License
- MIT (LICENSE file and README badge; the LICENSE copyright line still reads "[year] [fullname]")
- Installs on skills.sh
- 17,479 (Nov 2, 2026)
- skills.sh audits
- Gen Agent Trust Hub pass, Socket warn, Snyk warn (Nov 2, 2026)
- Last repo commit
- Oct 31, 2026
What it includes
Writes a feedback file under .vod/feedbacks/
Runs md_io.py write-feedback to save VOD-YYYYMMDD-NNNN.md in the project folder. Repeated errors in the same session add to a counter, and at most 5 files are stored per session by default.
Try: “This is a bug: the OBS upload failed again. Report it.”
Runs code ·
CaptureRedacts tokens, passwords and keys in the file
vod_sanitize.py replaces patterns such as token, password, secret, AK or HW_ACCESS_KEY values and bare keys that start with sk-, AKIA or LTAI.
Write ·
SanitizeThe agent fills in error, intent, scenario and dialog
The agent adds the error stack, what you wanted to do, what you expected, the product name, the environment (platform, OS, session ID, Python version) and 3 to 5 dialog turns around the problem.
Write ·
ExtractFiles the feedback as a GitCode issue
Runs vod_deliver.py deliver --feedback-id <id>, which creates an issue in the repository set in assets/config.yaml, then marks the file delivered with update-status.
Runs code ·
DeliverInstalls AtomGit-GO and signs you in to GitCode by QR code
If delivery needs a login, the agent runs vod_install.sh, starts a local server on port 8080, shows a QR code, waits up to 60 seconds, and the token is saved in plaintext at ~/.atomcode/auth.toml. A devspace-connector MCP tool is tried first if present.
Runs code ·
Auto-loginTriggers capture on tool errors, rejections and report phrases
hooks/openclaw/hooks.yaml and hooks/hermes/hooks.yaml activate the skill on failed tool results, on Chinese phrases such as 不对 or 这个有bug, and on report phrases.
Read ·
HooksRuns every script through skill-quality-cli
SKILL.md says every script must run as skill-quality-cli run --skill-name huawei-cloud-vod-collector -- <command>. The CLI collects quality telemetry and is switched off with SKILL_QUALITY_REPORT=0.
Runs code ·
Telemetry wrapper
From the official docs
Permissions
High risk
It files your error and chat context as a GitCode issue and installs a telemetry command and a login tool. Read each feedback file before you deliver it, and treat the issue as public.
What it can do
- Read your data
- Create and change things
- Delete things
- Read-only mode
Watch out for
Your error and chat context are filed as an issue
The feedback holds the error stack, your intent and 3 to 5 dialog turns, and delivery posts it to the GitCode repo in assets/config.yaml. The sanitizer redacts only common secret patterns. Read the file first and treat the issue as public. github.com
It installs a telemetry command and wraps every script in it
ensure_cli.sh and install_cli.sh fetch skill-quality-cli from skillsapi.developer.myhuaweicloud.com and SKILL.md requires every script to run through it. Opt out with SKILL_QUALITY_REPORT=0. SKILL.md does not list what the command reports. github.com
Login installs another tool and stores a plaintext token
vod_install.sh clones https://gitcode.com/weixin_45218422/AtomGit-GO.git and installs its binaries. The token is saved in plaintext in ~/.atomcode/auth.toml (mode 0600), and SKILL.md notes the GitCode API puts the token in the URL query. github.com
Hooks fire on errors and on disagreement phrases
The OpenClaw hook fires when a tool call exits non-zero with an error, and both hook files fire on phrases such as 不对 or 反馈问题. They are provided for OpenClaw and Hermes, which read the YAML files. github.com
It runs a local server on port 8080 during login
The login flow starts a local server and calls http://localhost:8080/login/start, then stops it. Close other tools using that port first. github.com
Security advisories
No published security advisories as of Nov 2, 2026.
Technical details
| Turning on read-only | No read-only mode. Delivery posts data. |
|---|---|
| Talks to | gitcode.com (issue API under /api/v5 and the AtomGit-GO repo). skillsapi.developer.myhuaweicloud.com (skill-quality-cli manifest). localhost:8080 during login |
| Sign-in and keys | GitCode token saved in plaintext at ~/.atomcode/auth.toml with mode 0600, per SKILL.md. Anyone with access to the file can use it. |
| Scopes |
|
Questions
What does huawei-cloud-vod-collector do?
It collects bad experiences with Huawei Cloud tools into a feedback file, redacts secrets, and delivers it as a GitCode issue for Huawei. It is a feedback tool for Huawei, not something that helps you use their cloud.
What does it install on my machine?
ensure_cli.sh downloads skill-quality-cli from skillsapi.developer.myhuaweicloud.com into ~/.local/bin after a SHA256 check. If delivery needs a login, vod_install.sh installs AtomGit-GO into ~/.local/bin, from a prebuilt archive in a cloned repo or by building from source. Python needs pyyaml and tomli.
Where does my feedback go?
To an issue in the GitCode repository named in assets/config.yaml (https://gitcode.com/developer-skill/vod-skill.git). Our advice: read the file under .vod/feedbacks/ before you deliver it and treat the issue as public. The sanitizer redacts common token and key patterns, not everything.
Can I turn the telemetry off?
SKILL.md says to set SKILL_QUALITY_REPORT=0 to opt out of the quality reporting that skill-quality-cli does. The same file says the wrapper is mandatory for every script run.
Does it work in Claude Code, Codex or the Claude app?
The install box has Claude Code, Codex and standalone tabs. Huawei says only that the skills are compatible with mainstream AI Agents and names none, and it documents no Claude app upload. This skill runs command-line tools on your machine, so the box has no Claude app tab. The hook files exist for OpenClaw and Hermes only, so elsewhere the skill loads by its description.
Is huawei-cloud-vod-collector official?
It is in Huawei Cloud's own huaweicloud/huaweicloud-skills repo, which the README calls officially maintained Huawei Cloud Agent Skills.
Information
updated Nov 2, 2026
- Repository
- huaweicloud/huaweicloud-skills
- Folder
skills/devtools/common/huawei-cloud-vod-collector
Sources
Docs, issue trackers and discussions this page is built from.
Show all 11Hide
- huawei-cloud-vod-collector SKILL.md · github.com
- huaweicloud-skills README · github.com
- huaweicloud-skills LICENSE · github.com
- VoD collector config.yaml · github.com
- install_cli.sh · github.com
- ensure_cli.sh · github.com
- vod_install.sh · github.com
- vod_sanitize.py · github.com
- OpenClaw hooks · github.com
- Hermes hooks · github.com
- hooks-setup.md · github.com