Yibiao Tender Toolkit User Manual Maintenance Skill
On demand, create or update Markdown user manuals, capture and annotate screenshots, write professional v2 changelogs, and publish to the issue-wiki after confirmation.
Evidence shows the skill explicitly requires non-sandbox execution (require_escalated) and automatically reads admin credentials from .env.local, logs in and writes to remote, with implicit authorization (selecting task is considered consent). There is room for overreach, but data flow and permissions are disclosed, and no malicious behavior found. Deductions for excessive permissions and incomplete confirmation.
Scripts show rigorous logic, clear error handling and atomic writes, but not executed, depend on external Git and remote API; key-path reproduction unverified. Deduction due to static review limitation.
Scenarios and trigger conditions clearly described, good Chinese support, but no mention of network dependencies or unreachable overseas services, and capability boundaries incomplete. Deduction for insufficient boundary evidence.
Documentation well-structured with detailed references and scripts, but lacks version number, changelog, and clear maintenance responsibility statement; license not declared within skill. Deduction for incomplete governance info.
Core tasks (update manual and version log) theoretically complete, but output requires user confirmation and manual review, and no actual runtime verification. Deduction for static review and unverified output.
Only source code and documentation, no test suite or CI evidence, key claims cannot be independently verified. Deduction for single evidence type.
- This skill requires non-sandbox execution and auto-reads admin credentials; ensure a trusted environment and explicit user authorization scope.
- Server publication involves remote write operations; always execute after explicit user confirmation and verify plan_hash to prevent plan changes.
- Depends on GitHub Releases and yibiao.pro overseas services; may be inaccessible from mainland China, inform users in advance.
- Publisher is not verified by FollowSkills; identity unknown, verify trustworthiness before use.
What does this skill do, and when should you use it?
This skill is designed for maintainers of the Yibiao tender toolkit. It creates or updates Markdown user manuals, captures and annotates screenshots, writes professional v2 changelog entries based on official Git tags, and compares the target documents with the issue-wiki before publishing after user confirmation. The workflow is fixed: determine scope → generate or update → server comparison → publish after confirmation. The skill only executes tasks explicitly selected by the user and does not add extra planning, testing, or fallback steps. It reads project contract, development notes, screenshot annotation formats, changelog standards, and runs specific scripts.
Reads reference documents such as references/project-contract.md, client/开发说明.md, annotation-format.md, and changelog-standard.md. Inspects the relevant feature code, release UI, and existing business data. Uses PowerShell scripts capture-window.ps1 and annotate-screenshot.ps1 to capture and annotate screenshots. Runs changelog_tool.py inspect --json to inspect v2 tags and insert --apply --json to write changelog entries. Runs sync_manual.py --json for a read-only server preview, generates an operations list, and after user confirmation executes with --apply --expected-plan-hash to publish.
- A maintainer asks to update the manual for a specific feature; the skill modifies only the corresponding Markdown section and captures screenshots.
- A maintainer requests updating the v2 changelog; the skill inspects Git tags and writes new, improved, fixed, or adjusted entries for missing versions.
- A maintainer chooses 'Update everything'; the skill updates both the full manual and changelog, then proceeds to server comparison.
- A maintainer asks to only sync existing local docs; the skill skips server comparison and syncs local documents.
- A maintainer asks for a read-only check; the skill only inspects without modifying or syncing.
- A user says just 'update manual'; the skill asks whether to update a specific feature or the full manual.
What are this skill's strengths and limitations?
- Fixed workflow reduces decision fatigue, clearly defining four steps: scope, generate, compare, publish.
- Supports multiple task modes: specific feature, full manual, changelog, all updates, local-only sync, read-only check.
- Uses official v2 Git tags to filter out refactors, tests, and other non-user-visible changes, resulting in a more professional changelog.
- Screenshot capture and annotation are automated, saving manual effort.
- Server comparison provides a preview, and publishing only happens after confirmation, reducing risk of accidental changes.
- The skill is specific to the Yibiao tender toolkit and not generalizable.
- Depends on specific scripts and reference documents, requiring Python, PowerShell, Git, and network access.
- No test suite or validation steps are mentioned; any error stops immediately without automatic retries.
- Documentation does not cover all platforms (e.g., Linux) for development setup, so you may need to adapt.
- Server comparison and publishing require admin API permissions, which may not be available to all users.
How do you install this skill?
Clone the repository to your local machine. Ensure the fork includes the .agents/skills/yibiao-user-manual/ directory and SKILL.md. The skill depends on the references/ and scripts/ directories within the repo. Python 3 and PowerShell are required, and you need access to the GitHub repo and the issue-wiki service. Detailed installation steps are not provided in SKILL.md; refer to the README for development environment setup.
How do you use this skill?
To invoke the skill, use natural language in your agent, for example: 'Update the full manual' or 'Update the v2 changelog'. The skill will follow the workflow: determine scope, generate or update, server comparison (unless 'local only' is chosen), and publish after confirmation. Ensure you have logged in to the issue-wiki admin API (if server comparison is involved) and authorized reading of .env.local. Before publishing, the skill shows the operations list and only writes after your confirmation.