Public, harness-readable skills for building a company in Known one workflow at a time.
Each skill is useful by itself. Install another when the business needs it: existing records remain canonical, optional references become stronger, and the company model grows without requiring a big-bang setup.
| Order | Skill | What it establishes |
|---|---|---|
| 1 | company-foundation |
Company profile, objectives, and operating facts |
| 2 | work-management |
Initiatives, tasks, decisions, and risks |
| 3 | relationship-management |
Organizations, contacts, interactions, and introductions |
| 4 | customer-development |
Customer hypotheses, opportunities, discovery, and pilots |
| 5 | product-development |
Feedback, product initiatives, and experiments |
| 6 | finance-and-runway |
Budgets, cash/runway snapshots, and financial assumptions |
| 7 | fundraising |
Rounds, investor prospects, materials, and introductions |
| 8 | hiring |
Roles, candidates, interviews, and hiring decisions |
This order is a recommendation, not a dependency graph. Start with any skill. Every cross-skill reference is optional and becomes available only when both concepts exist.
Point a compatible agent harness at one skill's SKILL.md. The skill instructs the harness to:
- discover or install the Known CLI;
- establish the correct authenticated organization and environment;
- inspect the existing wiki and its permissions;
- propose an adapted ontology before writing;
- wait for approval;
- create pages, attach JSON Schemas, and optionally add synthetic bootstrap records through the existing Known CLI;
- verify and report the resulting model.
The skills do not contain an installer and do not add commands to Known. The harness reasons; the Known CLI reads and writes.
Install the repository as a Git package:
pi install git:github.com/future-unknown/skillsThen invoke a skill explicitly when desired:
/skill:company-foundation Review my current Known wiki and propose the smallest useful company foundation. Do not write until I approve.
Other Agent Skills-compatible harnesses can load a skill directory directly.
- Schema-backed keyed records own the current state of the business.
- Every keyed-record update creates record history; inspect it with record history/diff or historical table reads instead of duplicating old state.
- Operationally meaningful events can be immutable structured records; they are not copies of obsolete state.
- Current table records appear in search, so the canonical collection is normally its own directory.
- Specialized indexes are schema-backed derived projections, never Markdown lists.
- A record's
contentcan carry focused Markdown and render as its page; do not create a separate detail page by default. - Separate Markdown pages remain right for long-form or cross-record narrative, strategy, process, evidence synthesis, and navigation.
- Source claims, assessments, workflow state, and canonical relationships stay separate.
- Unknown facts remain
nullor absent. A skill never invents intent, status, identity, affiliation, or commitment. - Skills inspect and reuse before creating. They do not establish a competing canonical collection without approval.
- Cross-skill IDs are nullable. Human-readable source claims are preserved when no canonical reference exists.
- Each monetary amount carries an explicit ISO 4217 currency.
- Sensitive personal, legal, health, credential, and payment data stays out of public or broadly visible wikis.
See references/data-modeling.md, references/ontology-principles.md, and references/composition.md.
references/ shared CLI, data-modeling, ontology, and composition guidance
skills/<name>/SKILL.md harness instructions
skills/<name>/references/ domain model and boundaries
skills/<name>/assets/ JSON Schemas and synthetic examples
scripts/validate.mjs repository validation used by CI
This is an initial public seed. Schemas are intentionally small and conservative. Treat live company data as canonical and review every proposed migration.
Read CONTRIBUTING.md. By contributing, you agree that your contribution is licensed under the repository's MIT License.