Quality AI infrastructure shouldn't be gatekept to the top 1%.

Write a capability once. Every AI surface gets it.

Scaffolde is a personal AI capability factory: skills, agents, workflows, and verification kept as versioned source, then projected into Claude Code, Codex, Hermes, OpenClaw, and the other runtimes where AI operators work. Upstream infrastructure, not another chat client.

  • One canonical source
  • 12 declared surfaces
  • Evidence-backed completion
Canonical source
One versioned repository: skills, agents, workflows, runtime services, verification.
Typed manifests
Validate and compose. Missing references stop projection.
12 declared surfaces
Claude CodeCodexHermesOpenClawPrime AgentPaperclip+ 6 more declared
Six primary executors, one secondary, one improvement worktree, four experimental. Declared, not a health guarantee.
Evidence returns
Doctor, readiness, tests, and Work Packets decide what earns its way back into source.
Projects intoClaude Code·Codex·Hermes·OpenClaw·Prime Agent·Paperclip·+ 6 more declared surfaces177 skills · nearly 40 agents · 235 workflows · 17 services · 51 jobs
How it works

One-way projection. Evidence back.

The repository is the factory; runtime surfaces are projections. A hand edit under ~/.claude or ~/.codex is drift, not authority: legitimate live fixes get promoted back into canonical source, then re-projected.

Canonical source
scaffolde-ai: capabilities, contracts, adapters, versioned.
Typed manifests
Declare what each surface receives. Missing references stop projection.
Sync engine
Writes managed output, snapshots file hashes, halts on drift.
Projected roots
~/.claude~/.codex.claude/skills+ other declared surfaces
Evidence back
Doctor, readiness, tests, Work Packets. Only evidence earns a change back into source.
1
Author canonical source. Capabilities, contracts, and adapters live in the repo first, versioned and reviewable, never as a one-off tweak inside a runtime directory.
2
Validate, compose, project. Typed manifests declare what each surface receives; missing references stop projection, and a sync engine writes managed output with drift detection.
3
Return evidence. Doctor, readiness, tests, and Work Packets decide whether an improvement earns its way back into source.
Read the full story
The flow

Author → validate → project → evidence.

Four steps turn a capability into versioned infrastructure every declared surface can use.

01 / AUTHOR

Capabilities as source

Skills, agents, workflows, and adapters are written into one versioned repository. The repository is the factory, so nothing exists only inside a runtime directory.

02 / VALIDATE

Typed manifests, fail-closed

Manifests declare what each surface receives, and validators check the references against the file set. Missing or inconsistent references stop projection.

03 / PROJECT

One-way sync with drift detection

The sync engine writes managed output into declared runtime roots, snapshotting file hashes so it halts before clobbering a hand-edited live file.

04 / EVIDENCE

Doctor, readiness, tests decide what returns

Surfaces report ready, degraded, attention, or blocked. Evidence from the runtime decides whether a live fix is promoted back into canonical source.

The evidence discipline

Done exists in writing before the work begins.

Scaffolde runs on Algorithm 8.4.0, a conjecture-and-refutation loop: the goal survives verbatim, done is a set of falsifiable claims, and each claim names the probe that could disprove it. “Should work” is forbidden.

  • The goal, verbatim. The operator’s request survives the loop unedited, so nothing is quietly reinterpreted into something easier to finish.
  • Falsifiable claims, written first. Done exists in writing before building, as a set of claims that could be shown false.
  • Right-modality evidence. A file claim needs a read, a deploy claim needs a live probe, a UI claim needs a real browser, never curl.
  • Mechanical gates. A VerificationGate hook grades completion claims against actual tool calls. Gaps block, and rewording never passes.
  • Honest grading. Some of this is mechanically enforced today (the VerificationGate hook, drift detection, fail-closed validators); the rest is honest self-attestation, and the repo grades each claim’s enforcement level by name.
Fail-closedEvidence-backed
Claim modalityEvidence required
File claimA read of the file itself
Deploy claimA live probe of the deployed target
UI claimA real browser, never curl
Completion claimTool calls that match it, or the gate blocks
No completion claim closes without tool evidence in the right modality.
What the factory owns

One factory. Every surface.

Everything below is versioned source in one repository, projected per-surface through typed manifests.

Skills

177 active skills across 26 declared domains, authored once as canonical source and projected wherever they are declared.

Agents

Nearly 40 named agents, including cross-vendor wrappers: Claude models plus Gemini, Grok, Qwen, DeepSeek, Kimi and more. Wrappers are agents in the factory, not verification coverage.

Workflows

235 multi-step workflows, versioned with the capabilities they call, so a change to either is a reviewable diff.

Surfaces

12 declared surfaces, six of them primary executors. A declaration says what the factory can project, not that a given machine is healthy.

Runtime services

17 declared services: voice, browser, observability, WhatsApp, knowledge, and more, each with its own contract in source.

Scheduled jobs

51 declared jobs: nightly ship, inbox checkers, memory health, secret scanning. Scheduled work is source too, not cron folklore.

Memory federation

Paperclip owns the durable backlog, gbrain synthesizes cross-surface knowledge with lineage, and human-authored intent wins conflicts.

Verification machinery

The VerificationGate hook, drift detection, doctor and readiness reporting, and 34 gate validators whose registry is asserted against the validator file set in CI, so a new validator fails the build until it is registered.

Vendored lines

Paperclip and gbrain are vendored upstreams with explicit patch boundaries. The factory carries them, it did not author them.

Factory counts

The factory, counted.

Counts come from the repo’s generated indexes and manifests, August 2026. Surfaces are declared, which says what the factory can project, not that every machine is healthy.

177Active skills
235Workflows
10,900+Tests green
51Scheduled jobs
12 Declared surfaces26 Domains13 Bundles17 Runtime services34 Gate validatorsNearly 40 Named agents
Market & why now

Every new agent runtime is another config directory.

Operators now run several agent runtimes at once, and each one keeps its own copy of the same capabilities. The work compounds in the runtime, where it cannot be versioned, and never in one place that outlives the tool.

Runtimes multiplied

Claude Code, Codex, Cursor, and a wave of agent CLIs each keep their own skills, prompts, and settings under their own root directory.

Capability got fragmented

The same skill gets rewritten per tool, drifts per machine, and dies with the session that produced it.

Hand edits don’t scale

A live tweak under ~/.claude is unversioned, unreviewed, and silently lost the next time the tool updates its own files.

The missing piece is upstream

A canonical factory that versions capability once and projects it everywhere, with drift detection holding the line and evidence deciding what returns to source.

Traction

A working system, honestly measured.

The status, stated plainly: private, single-operator, macOS, with v2.0.0 as a release candidate. Everything below is checkable in the repository.

Working systemPrivateScaffolde runs the founder’s daily AI work today: 177 active skills, 51 scheduled jobs, and 12 declared surfaces, maintained as a private single-operator deployment on macOS.Single operator
Engineering evidenceGreen10,900+ maintainer-health tests passing across 991 files, 34 gate validators in a registry CI-asserted against the validator file set, and a release gate that fails closed. v1.0 and v1.1 tagged; v2.0.0 is a release candidate.Green
The raisePre-seedRaising to productize the factory beyond a single operator and take the founder full-time.OpenSee the thesis
Scaffolde is the factory. Runtime surfaces are projections.
The root invariant

See the factory model.

No form, no sales call. Read how canonical source becomes a projected surface, and what evidence it takes to change the factory.