Back to search
dsh-founder-skills
SkillGitHub
The Founder pack for DeepSeek Harness: eleven business skills plus one chained entry point that runs a complete business-idea assessment (Quick 20 buyers, Full 100).
Safety scan
A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.
Badge for your README
Markdown
[](/listings/boldandigital--dsh-founder-skills)HTML
<a href="/listings/boldandigital--dsh-founder-skills"><img src="/api/listings/boldandigital--dsh-founder-skills/badge.svg" alt="findsafeskills safety rating" /></a>Install
i
Install instructions from the README
As written by the repo's author. findsafeskills has not verified these steps.
git clone <this-repo> && cd founder-dsh ./install.sh # into ~/.dsh/skills ./install.sh --profile work # into ~/.dsh/profiles/work/skills ./install.sh --uninstall # remove it
No plugin, no build step, no network. DSH's skill-filesystem provider discovers each `skills/<name>/SKILL.md` automatically.
What was scanned
The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.
README.md · README · 172 lines
- # Founder for DeepSeek Harness
- Eleven specialist skills and one entry point that runs **all of them**, end to
- end, on one business idea. Give it an idea; get a business plan with a verdict
- that is computed from the numbers rather than written by hand.
- Ported from [Jakeschincariol/founder-skill](https://github.com/Jakeschincariol/founder-skill)
- (MIT) and adapted to DeepSeek Harness.
- ## Install
- ```bash
- git clone <this-repo> && cd founder-dsh
- ./install.sh # into ~/.dsh/skills
- ./install.sh --profile work # into ~/.dsh/profiles/work/skills
- ./install.sh --uninstall # remove it
- ```
- No plugin, no build step, no network. DSH's skill-filesystem provider discovers
- each `skills/<name>/SKILL.md` automatically.
- ## Use it
- ```
- test this idea: a mobile dog-wash service in Lisbon
- ```
- That one request runs the whole chain. The `founder` skill creates an isolated
- run directory, drives each stage in order, and finishes with the compiled plan.
- | say | you get |
- | --- | --- |
- | "test this idea" | Quick — the full pack, 20 simulated buyers |
- | "test this idea, full run" | Full — the same pack, 100 simulated buyers |
- | "how much should I charge for…" | just `founder-pricing` |
- | "would people buy this…" | just `founder-consumer` |
- **Quick and Full are the same product.** Every stage runs identically; only the
- panel size changes. A Quick run is a real business plan with less statistical
- power, not a stripped-down one. Both write the same files.
- The practical difference is time. The harness allows eight active child agents
- at once, so the panel runs in waves of six to eight:
- | | buyers | waves | wall time |
- | --- | --- | --- | --- |
- | Quick | 20 | 2–3 | an hour or two |
- | Full | 100 | 10–13 | several hours |
- Expect a few relaunches in Full. A child that reports success without writing
- its answer file has not answered — `panel.py check` names every buyer still
- missing, which is what the loop runs on.
- ## What comes out
- One directory per idea, so two ideas never share state:
- ```
- runs/my-idea/founder/
- ├── idea.md board.md competitors.md competitors.csv
- ├── customer.json pitch.md panel/{panel.json,personas.json,briefs/,answers/,results.md}
- ├── pricing.md pricing-curve.md offer.md numbers.json
- ├── cfo.md cfo-sources.md marketing.md brand.md
- ├── ops.md launch.md
- └── summary.md business-plan.md one-pager.md
- ```
- `business-plan.md` opens with a verdict — **Profitable**, **Not yet** or
- **Incomplete** — computed by `compile.py` from four checks:
- 1. each unit earns money before fixed costs,
- 2. year 1 makes an operating profit,
- 3. break-even fits inside capacity,
- 4. enough simulated buyers buy (default bar 25%).
- A failing check is never rounded into a passing one, and the plan lists which
- stages are still missing.
- ## The chain
- ```
- founder-board → founder-competitors → founder-consumer → founder-pricing
- → founder-offer → founder-cfo → founder-marketing → founder-brand
- → founder-ops → founder-launch → founder-plan
- ```
- The early stages are cheap and decisive; the late ones are expensive and
- committal. Pricing comes after the panel, because the panel's objections set
- the price. The CFO comes after the price, because the price is the CFO's
- biggest input.
- Each specialist still works alone. `founder-plan` reads whatever the others
- wrote and tells you what is missing.
- ## What is honest about it
- - **The panel is simulated.** It surfaces objections and price reactions. It
- never proves demand. Only a real test — a pre-sale, a paid order — proves
- demand, and `founder-launch` puts a real test first, with a number written
- down before it runs.
- - **No invented facts.** Every competitor price, licence rule and supplier
- price carries a source and a date. Where evidence is missing, the stage says
- so instead of filling the gap.
- - **The verdict is computed.** `compile.py` decides. No stage may soften a
- failing check.
- ## Currency
- Set `"currency"` in `numbers.json` and the money renders in it everywhere —
- unit economics, the compiled plan, the panel's income bands:
- ```json
- { "currency": "R$", "price": 24.00, "unit": "order" }
- ```
- Supported by `unit_economics.py`, `compile.py`, `van_westendorp.py` and
- `panel.py`. The default is `$`, and the upstream test suite passes unchanged.
- ## Layout
- ```
- founder-dsh/
- ├── install.sh installer
- ├── package.json bundle manifest for `dsh plugin install`
- ├── cordis.patch.yml patch layer for the bundle route
- ├── LICENSE MIT (upstream)
- ├── README.md
- └── skills/
- ├── founder/ the entry point: SKILL.md, workflow.md, run.py
- ├── founder-board/ SKILL.md, lenses.md
- ├── founder-brand/ SKILL.md
- ├── founder-cfo/ SKILL.md, unit_economics.py, example.json
- ├── founder-competitors/SKILL.md
- ├── founder-consumer/ SKILL.md, panel.py, customer.example.json
- ├── founder-launch/ SKILL.md
- ├── founder-marketing/ SKILL.md
- ├── founder-offer/ SKILL.md
- ├── founder-ops/ SKILL.md
- ├── founder-plan/ SKILL.md, compile.py
- └── founder-pricing/ SKILL.md, van_westendorp.py
- ```
- ## Adapting it to DSH
- The upstream pack is written for Claude Code. The port changed:
- - **No `${CLAUDE_SKILL_DIR}`.** Each skill refers to its own directory as
- `$SKILLDIR`, the path DeepSeek Harness prints in the skill header.
- - **No `/founder-x` slash commands.** Skills refer to each other by name.
- - **No `$ARGUMENTS` or `argument-hint`.** Instructions are plain prose.
- - **The Claude agent tool is the DSH `subagent` tool.**
- - **New: the `founder` entry point.** The upstream pack expects a human to
- invoke each skill in turn. That is the wrong shape for an agent, so the
- entry point drives the chain itself and never stops at a stage's printed
- "Next:" pointer.
- - **New: `run.py`.** Creates and inspects isolated run directories, so two
- ideas never share `founder/`, and a resumed run knows what is left. It also
- validates a customer profile before the panel is set up, so a schema mistake
- surfaces immediately instead of halfway through a run:
- ```bash
- python3 run.py new --name my-idea --root runs --mode quick --currency '€'
- python3 run.py check-customer --customer founder/customer.json
- python3 run.py status --dir runs/my-idea
- ```
- - **New: currency support** in the four scripts that print money.
- ## Not legal, financial or tax advice
- The licence and permits section points at official pages and says plainly that
- rules differ by place and change. Confirm each one with the authority.
Ratings and comments
Sign in to rate this skill.
Sign inFree to use. If it’s helped, pass it on.