DSH Career Skills

Three focused job-search skills for DeepSeek Harness, plus a router:

• job-fit — decide whether a posting is worth applying to. Liveness and dealbreaker gates, a 1–5 score, must-have mapping, level/comp, red flags.
• resume-tailor — tailor a CV, resume, or cover letter to a JD without inventing facts. ATS rules, evidence mapping, [NEED: ...] for gaps.
• interview-prep — company research, a STAR+R story bank, likely questions mapped to stories, questions to ask, a seven-day plan.
• career — routes a request to the right one of the three.

Why this is better than a raw port

The source methodology is career-ops (MIT). This is not its 600-script pipeline copied over. It is a native DSH adaptation:

• No Node/Python scripts, no Playwright, no npx init, no config file to edit.
• Four skills instead of a full CLI project; the router keeps the catalog small.
• Uses DSH tools directly: read, write, web_fetch, ask_user_question.
• Local-first, no auto-submit, no email sending, no telemetry.
• Every output is a draft the user reviews.

Install

DSH bundle

From a local checkout, install the folder through Plugins → Add plugin, or with the plugin manager. After the package is published to npm:

dsh plugin --profile web add dsh-career-skills

Copy into the skills root

./install.sh

or:

mkdir -p "${DSH_HOME:-$HOME/.dsh}/skills"
cp -R skills/* "${DSH_HOME:-$HOME/.dsh}/skills/"

Use

Ask normally:

• "Evaluate this job against my CV: "
• "Tailor my CV to this role: "
• "I have an interview at for ; prepare me."

Or start with the router: "Use the career skill to decide which skill fits this."

Put the CV in cv.md in the workspace and the target profile in .agents/career-profile.md (roles, locations, compensation floor, hard dealbreakers).

What it does not do

• It does not scan job boards or auto-apply. Paste a posting or a URL.
• It does not generate PDFs unless a converter is already available.
• It does not know the user's salary history or private facts; it asks or marks [NEED: ...].
• It is not a recruiter and does not guarantee outcomes.

Attribution

The evaluation structure and workflow are inspired by career-ops, MIT. No career-ops code is copied; the skill texts are original DSH-native adaptations. See NOTICE.md.

License

MIT, see LICENSE.