xuzilong-skills

KO | EN

gitlite — search

#javascript #python #hacktoberfest #react #ai #typescript #llm #go #golang #android #machine-learning #rust #deep-learning #linux

xuzilong-skills

★ 18 Open GitHub ↗

This is a skill repository for AI agent. And it is developed especially for Economic and Management academic research.

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# xuzilong-skills

Design-first, governance-oriented skill set for academic AI workflows.

[中文说明](README.zh-CN.md)

## Why This Repository Exists

This repository is not a generic prompt collection.  
It defines **operational contracts** between skills so research workflows remain:

- reproducible by default
- auditable at phase boundaries
- explicit about ownership (who writes what, who only records)

The core design target is stable execution across `clean` and `analysis` stages, with strict separation between:

- research pipeline control (`xuzilong-academic`)
- task log governance (`xuzilong-task-tracking`)

## Design Principles

- **Hard gates over soft reminders**: required artifacts are enforced, not suggested.
- **Single source of truth**: manuscript numbers must come from result interfaces.
- **Protocol-driven collaboration**: inter-skill updates use structured YAML blocks.
- **Clear boundaries**: task tracking records status; it does not alter research logic.
- **Script-bound intermediates**: intermediate files map to script names, not ad-hoc notes.

## Skill Architecture

- `xuzilong-academic/`
- `xuzilong-task-tracking/`

### `xuzilong-academic`

Controls academic pipeline constraints and interfaces:

- stage gating for `clean` / `analysis`
- required intermediate artifacts in `Program/*/temp`
- manuscript-number consistency checks
- YAML handshake output to task-tracking

### `xuzilong-task-tracking`

Controls task-state recording only:

- reads/validates YAML update blocks
- appends updates to `Tasks/task-current.md` (append-only)
- refuses invalid state promotion (e.g., no `done` on invalid blocks)
- does not rewrite script paths or technical conclusions

## Inter-Skill Handshake (Concept)

Top-level task envelope + script list:

- `task_id`
- `task_status`
- `scripts[]`
- `updated_at`

Per script:

- `script_name`
- `script_type` (`clean` / `analysis`)
- `status`
- `artifacts`
- `latest_result`
- `acceptance_check`

`task_status` must be strictly aggregated from script statuses.

## Runtime Context Model

- `Tasks/outline.md` can serve as always-readable research context.
- Editing `Tasks/outline.md` is restricted to explicit task-tracking use.
- This supports broad context visibility without accidental drift.

## Compatibility

Primary target:

- Codex / OpenCode style skill workflow (`AGENTS.md` + `SKILL.md`)

Also compatible with:

- Claude Code style skill usage (`SKILL.md` conventions)

## Access Control (GitHub)

Yes, you can keep this repository visible to **only specified people**:

- set repository to **Private**
- invite specific collaborators by GitHub username/email
- optionally assign least-privilege roles (`Read`, `Triage`, `Write`, etc.)

## License

`TBD` (set before public release)

xuzilong-skills

Explore Similar Repositories

Was this content helpful?

Recent Feedback

Download README