From career-hub
Interview-driven enrichment of a career hub. Given a category (foundations | roles | projects | capabilities) or a menu if no argument, reads existing stubs, drafts from known content, and asks the user only gap questions. Edits files in place, bumps last_updated, flips status from draft to active when substantively complete. Offers .private.md confidential siblings when candid content surfaces. Logs deferrals to g.backlog/ as one file per item.
How this skill is triggered — by the user, by Claude, or both
Slash command
/career-hub:career-hub-buildThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Enrich a career hub via conversational interview, one category at a time.
Enrich a career hub via conversational interview, one category at a time.
Typical prompts:
Hub-first, rendering-second. The hub is the source of truth; CVs, LinkedIn, and GitHub profile pages are views onto it. Every public claim traces to an entry in the hub.
Draft-from-known-content, ask-only-gap-questions. When build enriches a stub, Claude drafts from existing hub content and cross-references, and asks the user only gap questions — never redundant questions. Fast, low-friction enrichment.
Every claim verifiable. No unverifiable adjectives ("expert", "pioneered"). Scale numbers and verification methods live in a.foundations/biographical-facts.md.
cwd is a compliant hub (see ../references/hub-detection.md). Refuse otherwise with a pointer to career-hub-init.Hub detected at <cwd>.
- a.foundations: <4 files, X complete / Y stubs>
- b.history: <N roles, X active / Y draft>
- c.projects: <N projects, X active / Y draft>
- d.capabilities: <N cards, X active / Y skeleton>
- g.backlog: <N open threads>
Proceed with build <arg>? (y/n)
On explicit y, proceed.
If invoked with a category argument (build roles, build projects, build capabilities, build foundations):
status: draft file in chronological order).If invoked bare (build):
Which category?
1. foundations (4 files, 2 stubs)
2. roles (12 files, 5 stubs)
3. projects (9 files, 3 stubs)
4. capabilities (6 skeletons)
For each file selected:
a.foundations/biographical-facts.md).last_updated in frontmatter.status: draft → active if the unit is now substantively complete.<filename>.private.md sibling for candid notes?" — on yes, create the sibling with the confidential template.g.backlog/<slug>.md entry per the backlog item template, and register it in g.backlog/README.md index.After editing any role / project / capability file, also update:
README.md — if the file's row caption changed (scope, active/draft status, headline).a.foundations/biographical-facts.md — if a numerical claim, tenure, or scope has changed.One commit per completed unit, OR one commit per category pass — operator's choice, default to one commit per unit.
Commit message format:
content(<category>): enrich <slug> — <brief summary>
Example:
content(b.history): enrich 2020-2021-nab-ad-optimisation — scope, outcomes, team structure
Draft from known content. Ask only gap questions.
The skill's differentiator. Don't ask what you can infer. Don't ask what cross-references already state. Don't run a fixed question list.
For each unit, after reading, your first message to the user is:
**<file name>** — draft from known content:
[short proposed fill-in for each body section, marked where gaps exist]
Gap questions:
1. <specific thing you can't infer>
2. <specific thing you can't infer>
3. <specific thing you can't infer>
Or skip ahead and answer in one message if easy.
Keep gap questions to 3-5 per unit on a quick pass. Offer deep-dive mode on request for high-CV-weight items.
Default is quick pass — fast gap-filling, target complete unit in one exchange where possible.
Deep dive on request — more thorough probing, additional questions about politics / proprietary details (offer .private.md sibling), follow-up on stories that would make good case studies.
cwd not a hub (per hub-detection.md).status: active unless user explicitly asks to re-interview.publish's job).Provides CDSS development patterns for drug interaction checking, dose validation, clinical scoring (NEWS2, qSOFA), and alert classification integrated into EMR workflows.
npx claudepluginhub hiivmind/career-hub