ASu-skills: Nine AI Skills for the Full Job Search Pipeline

Nine Skills for the Full Job Search Pipeline

ASu-skills is structured as a plugin package, not a single monolithic tool. After installation, it registers nine independently callable skills. Each skill has a defined input, a defined output, and a specific stage of the job search it covers.

The nine skills are: /contributor, which finds real open-source issues to fix and manages the PR lifecycle; /evidence-recap, which converts AI programming session logs into structured nine-segment evidence chains; /project-guide, which generates reading plans, tutorials, and interview talking points for a given codebase; /great-resume, which rewrites experience against a job description and generates outreach text for recruiters; /make-resume, which produces an editable HTML resume from a set of templates; /job-match, which compares a job description to an applicant's experience and flags hard requirements and evidence gaps; /job-apply, which connects to a browser and fills out job application forms; /interview, which generates interview questions and conducts practice sessions; and /offer, which tracks application status, emails, and deadlines.

The README describes two design principles behind this structure. First, each skill can be used in isolation, so a developer who only needs to build a resume does not need to engage the full pipeline. Second, the skills share a common set of directories, skills/, assets/, and references/, so context built in one skill carries forward when another skill picks it up.

Installing ASu-skills on Claude Code, Codex, and Other Platforms

ASu-skills supports seven platforms. Claude Code and Codex receive full native support. Cursor, OpenCode, WorkBuddy, TraeWork, and Qoder are supported through bridge installations with separate setup guides.

For Claude Code, the install sequence from the README is:

/plugin marketplace add Hisn00w/ASu-skills
/plugin install asu-skills@asu

After those two commands, the README instructs the user to run:

/reload-plugins

Uninstalling reverses the process:

/plugin uninstall asu-skills

For Codex, the README instructs sending the GitHub repository link to Codex with a message to install it, then opening a new conversation and selecting skills from the / menu.

For TraeWork, the README specifies copying the repository into the directory `~/.trae-cn/plugins/<publisher>/asu-skills/<version>/` and restarting TraeWork.

The package requires Node.js 22.19.0 or later, or Node.js 24.0.0 or later. The package version documented in the repository is 0.4.0.

The README states that all platforms share the same skills/, assets/, and references/ directories. The catalog of available skills is stored in skills.registry.json at the repository root. A synchronization script, accessible as `npm run sync:skills`, regenerates the registry from the individual skill files. The CI pipeline runs the same script with a `--check` flag to verify that the registry matches the actual skill files, catching any drift between the two.

How the Nine Skills Interconnect

The README presents a set of recommended entry points based on a developer's current situation. A developer with no internship experience and no verifiable projects is directed to start with /contributor, which finds a real GitHub issue, walks through the change, and submits the PR under the developer's own account. Once the PR is merged, /contributor hands its output to /great-resume for incorporation into resume bullet points.

A developer who has used an AI coding assistant to build a project but cannot clearly separate their own contributions from AI-generated code is directed to /evidence-recap. That skill takes conversation logs and delivery records and organizes them into nine types of evidence: personal decisions, collaborative actions, delivery milestones, outcome data, and so on. The output is meant to be defensible in an interview where a recruiter challenges whether the developer actually understands the work.

For developers preparing to apply, the typical sequence is /project-guide to build interview talking points from the codebase, then /great-resume to align the resume to a specific job description, then /make-resume to render the result as an editable HTML file with PDF export.

The /job-match skill fits between /great-resume and /job-apply. It takes a job description and the candidate's documented experience and outputs three categories: requirements already matched, requirements that could be addressed with better phrasing, and true gaps where the candidate lacks the underlying experience.

The /offer skill operates independently of the rest, tracking applications, test dates, interview rounds, and incoming emails without modifying the resume or the job search strategy.

The /make-resume Skill and Its Template System

The README describes /make-resume as responsible for all resume file delivery. It defaults to the ASu single-column high-density technical template when no other template is specified, but the user can specify any of 18 Chinese HTML templates included in the repository, provide a local HTML file, or upload a screenshot of a target design for the skill to analyze and approximate.

The generated file is described as a genuinely editable HTML document, not a static screenshot embedded in a page. The README lists the editing capabilities available in a browser: text, photographs, fonts, colors, and bold formatting. A local-font option reads system-installed fonts using the Font Access API in Chrome 103 or later, with a browser permission prompt. A font-import option loads TTF, OTF, WOFF, or WOFF2 files from the local file system.

The skill also supports exporting a LaTeX .tex source file, with three layout options: the default ASu layout, a compact single-column layout, and an academic layout. The README notes that this option must be explicitly requested and is intended for editing in environments like Overleaf.

The /make-resume skill is distinct from /great-resume. The former handles formatting and file output; the latter handles content rewriting against a job description. The README recommends running /great-resume first to finalize the content and then /make-resume to produce the output file.

Where ASu-skills Has Boundaries

The README states an explicit constraint on /great-resume: when the input material is insufficient, the skill produces a draft with [pending] markers rather than inventing job titles, companies, technology stacks, or performance numbers. This is described as an intentional design choice, not a limitation to be worked around.

The /contributor skill has a parallel constraint. It tracks PR status as submitted or in review. It uses stronger language, such as merged or adopted, only when the GitHub page shows a merged status and the project documentation explicitly defines the scope that was addressed. This means a developer who submits a PR to a large project but does not receive a merge confirmation cannot claim that contribution in stronger terms.

The /job-apply skill requires a logged-in browser session to fill out application forms. It states it will stop before the final submission step and wait for the user to review and confirm. The README describes this as a deliberate checkpoint, not a deficiency.

Across all nine skills, the package targets Chinese-language job seekers applying to Chinese companies. The resume templates use Chinese layouts, the /offer skill tracks application statuses common in the Chinese campus recruiting calendar, and the /great-resume skill generates outreach messages for Chinese recruitment platforms like Boss Zhipin. Developers applying to companies outside China would need to adapt this infrastructure substantially.

Comparing with Ad-Hoc LLM Prompting

A developer with access to an AI coding assistant can accomplish most of what ASu-skills provides through direct conversation, without installing a plugin. The difference is consistency and transfer of context. With unstructured prompting, each session starts from scratch: the model does not know which open-source PR was submitted last week, what evidence was recapped in a previous session, or what evidence gaps /job-match identified for a specific role.

ASu-skills addresses this by storing intermediate outputs in structured files under the skills/, assets/, and references/ directories. When /contributor hands a merged PR to /great-resume, it does so through a shared file path that persists across sessions. When /evidence-recap produces an evidence chain, that output is available to /project-guide in a subsequent session.

A dedicated career coaching platform such as a commercial resume service would provide a visual interface and professional human review, which ASu-skills does not offer. The tradeoff is that a human-reviewed service does not integrate with a coding assistant's session history and cannot, for example, extract verifiable contributions from a GitHub commit log in the same automated way.

Maintenance and License

The last push to the repository was on 2026-09-24, four days before this article was written. The package is at version 0.4.0 and carries an MIT license, which permits commercial use, modification, and redistribution without restriction beyond attribution.

The README includes a CONTRIBUTING.md reference covering local validation, test commands, and PR submission requirements for developers who want to add skills or templates. The test suite runs with `npm test` for Node.js tests and `python3 -m unittest discover -s tests -t . -v` for Python validation scripts. The skill registry sync check is part of the CI pipeline and will fail if the registry is out of date.

Editorial conclusion

ASu-skills is a good fit for Chinese-language job seekers who already work inside Claude Code, Codex, or TraeWork and want a structured pipeline rather than improvised prompting. The strongest case for it is /evidence-recap, which converts AI programming session logs into traceable evidence chains, and /job-apply, which automates form filling against live job postings. Developers outside those environments, or those looking for a general resume builder that runs in a browser, will need to check whether a bridge installation for Cursor or OpenCode meets their specific workflow before committing to it.

Frequently asked questions

Official sources

1. Hisn00w/ASu-skills on GitHub
2. Issues
3. License: MIT
4. README