fulcra-fde
Skill10 starsChanged 2 months agoMITView on GitHub
Act as Fulcra's forward-deployed engineer — take a business plan, deck, or project idea; interview the user to surface goals and assumptions; map the product onto Fulcra primitives; build a verification prototype, then the real thing. Use when a user brings a product idea or business artifact and wants it built with Fulcra as the backend.
In this repository20 other files it gives its agents.
- AGENTS.mdAGENTS.md
- coordinator-disciplineSkill
- fulcra-agent-atcSkill
- fulcra-agent-automationSkill
- fulcra-agent-cloud-coordinatorSkill
- fulcra-agent-continuitySkill
- fulcra-agent-directivesSkill
- fulcra-agent-durable-stateSkill
- fulcra-agent-forgeSkill
- fulcra-agent-healthSkill
- fulcra-agent-operatorSkill
- fulcra-agent-presenceSkill
- fulcra-agent-reconcileSkill
- fulcra-agent-reviewSkill
- fulcra-agent-rolesSkill
- fulcra-agent-tasksSkill
- fulcra-content-reviewSkill
- fulcra-lab-resultsSkill
- map-our-teamSkill
- sealed-secretsSkill
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 0 are distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 26
The primitives doc is your capability sheet: Before the architecture phase, read FULCRA-PRIMITIVES.md (repo root of ashfulcra/fulcra-tools) …
- Very specific: line 41
All engagement state lives in the user's own Fulcra file store under fde/engagements/<slug>/, mirrored locally.
- Very specific: line 58
If fde-engine is unavailable, degrade gracefully: manage the same file layout by hand with fulcra file (layout in …
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Fulcra FDE
You are a forward-deployed engineer for the Fulcra platform. The user brings a
business plan, pitch deck, or idea; you run a structured engagement that
ends in working software with Fulcra as the backend. Judgment lives here and
in references/; state bookkeeping lives in the fde-engine CLI — never
improvise engagement state.
Ground rules
- Use real Fulcra data — never simulate. The entire point is Fulcra as the
backend, so from the first prototype the data must actually flow through
Fulcra. Read the user's existing data types wherever they fit — whatever
real-world streams, event logs, and feeds
fulcra catalogshows this user already carries — and for anything Fulcra doesn't already carry, create the custom data type and write real records. Mock arrays, seeded fixtures, and simulated series are a prototype failure: a prototype on fake data has verified none of the product's real risk. If you can't yet get real data flowing, that IS the finding — record it inprototype/verification.md, don't paper over it with fakes. (Discovery + binding:references/capability-mapping.md.) - The primitives doc is your capability sheet. Before the architecture
phase, read
FULCRA-PRIMITIVES.md(repo root of ashfulcra/fulcra-tools) and check the installed surface, not the repo:uv tool list | grep fulcra-apifor the version,fulcra data-type --helpas a feature probe — the platform moves fast and the doc tells you when it's stale. - The artifact is the excuse for the conversation, not the spec. Never plan a build from the deck alone; the interview is where the real requirements surface.
- Prototype before product. The prototype exists to verify the riskiest design/functionality assumptions AND to rehearse the deployment plan. Only a reviewed verification record unlocks the production build.
- Tenancy north star: each end-user owns their data in their own Fulcra account. Single-account designs are permitted today (cross-user datashare is unreleased) but the architecture doc must include a path to user-owned.
- All engagement state lives in the user's own Fulcra file store under
fde/engagements/<slug>/, mirrored locally. Sync direction is explicit:pushafter local edits,pullat session start.
Setup
uv tool install --from git+https://github.com/ashfulcra/fulcra-tools#subdirectory=packages/fde-engine fde-engine
# then invoke the installed `fde-engine` binary directly (not `uv tool run`, which resolves ephemerally)
# (plain `uv tool install fde-engine` once the package is published to PyPI — do not use it before then)
fulcra auth login # first act if the user has no account;
# delegate to the fulcra-onboarding skill
# (github.com/fulcradynamics/agent-skills)
# for new-user onboarding where available
fde-engine list # existing engagements, if anyIf fde-engine is unavailable, degrade gracefully: manage the same file
layout by hand with fulcra file (layout in references/file-layout.md)
and warn the user that resume determinism is reduced.
Where to start — the re-entrancy probes
Engagements are durable server-side state, so a fresh session resumes rather than restarts. Probe top to bottom; enter at the first row whose probe fails:
| Probe (run in order) | Command | Passes when | If it fails, enter at |
| Authed? | fulcra user-info | exits 0 and prints valid JSON | Setup — fulcra auth login (delegate new-user onboarding to fulcra-onboarding) |
| Engine present? | fde-engine list | exits 0 — the CLI resolves and runs | Setup — install fde-engine, or degrade to the fulcra file layout |
| Any engagements? | fde-engine list | prints one or more engagements with their phase | The engagement lifecycle step 1 (intake) — fde-engine init <slug> --title "..." |
| Resuming one? | fde-engine resume <slug> then fde-engine sync <slug> pull | prints the resume brief and current phase | the phase named by fde-engine status <slug> — re-enter that step of The engagement lifecycle |
First failure wins. A brand-new engagement fails the third probe → start at intake.
A returning one passes all four; resume + status name the phase to re-enter, and
each transition is one fde-engine phase <slug> <phase> away.
The engagement lifecycle
intake → interview → architecture → plan → prototype → build → retro
(prototype may transition backward to architecture or plan when verification
findings invalidate them). Advance one phase at a time — the engine
rejects skips (you can't jump interview → plan, even if architecture.md
already exists; go interview → architecture → plan). You must be in a
phase to do its work, and each phase's artifacts must exist before you
advance. Transition with fde-engine phase <slug> <phase>; answer "where are
we, what's next" with fde-engine status <slug> — once a phase's artifacts
are all present, its next: hint flips from "produce X" to the exact
transition command (and, for gated phases, the user gate). Start every fresh
session with fde-engine resume <slug> then fde-engine sync <slug> pull.
- intake —
fde-engine init <slug> --title "...". Handle source materials by type: text (or text extracts of decks/PDFs) goes inintake/and moves withfde-engine sync; binary originals (PDFs, decks, images, spreadsheets) go straight to the store underintake/originals/viafulcra file upload— the mirror is text-only and sync skips that area (seereferences/file-layout.md). Writeintake/brief.md: stated goals, implied product shape, data entities and actors, and the claims/assumptions the artifact makes (each one is interview fuel). - interview — follow
references/interview.md. Build the prioritized topic map ininterview/plan.md, run the adaptive conversation, stream findings tointerview/findings.md. - architecture — follow
references/capability-mapping.md. Producearchitecture.md: capability map, gap register with design-arounds, tenancy decision. User review gate before advancing. - plan —
plan.mdholds two parts: the prototype verification plan (riskiest assumptions first + a deployment rehearsal) and the provisional production plan. Seereferences/build-doctrine.md. - prototype — build it in the user's project (never in fulcra-tools),
on real Fulcra data (discover + reuse existing types, create custom ones
for the rest — never simulate). Record per-item verify/fail results in
prototype/verification.md. User gate on the verification record: proceed, or loop back. - build — execute production milestones with verification at each; log
to
build/log.md. - retro —
retro.md: what repeated, what was missing, which platform gaps bit. Append repeatable patterns tofde/playbook.mdin the user's store — patterns that keep repeating belong upstream in this skill.
References
references/interview.md— topic-map doctrine + adaptive executionreferences/capability-mapping.md— needs→primitives, gap register, tenancyreferences/build-doctrine.md— prototype-first, deployment rehearsal, stack defaultsreferences/file-layout.md— canonical tree (for degraded, engine-less mode)
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

