the pack registry
A pack is a Git repo of workflows and capabilities — a YAML library, nothing to compile. A tool is an MCP server a pack spawns to do real work. Point your gateway at a pack, andpraxec doctor provisions its required tools for you, with your consent.
packs vs. tools
Packs and tools are different kinds of thing, and they install differently:
repos: entry. They carry the workflows, capabilities, skills, and scripts — the governed behavior. Nothing gets compiled or installed; the kernel reads them at load and namespaces them.kind: mcp capabilities. Each tool resolves through an ordered provider chain — Docker image first, then a release binary, thencargo install — and praxec doctor detects what's missing and offers the provision command before running anything.So the flow is: load a pack, run praxec doctor, and it reports which required tools aren't present yet — then provisions them (with consent) down the chain. You never hand-wire the servers a pack depends on.
workflow packs
Capability and orchestrator libraries, loaded into a gateway viarepos:. Coding is the first use case, not the only one.
The canonical SWE-lifecycle library — lifecycle orchestrators (flow.add-feature, flow.bugfix-from-error-log, flow.safe-refactor, flow.triage-issue, flow.evidence-driven-convergence) and the capabilities, skills, and scripts they compose from. The pack most engineers running a coding agent want first.
repos:
- path: ./packs/cognitive-architecturesThe MAX-tier UX factory layered on cognitive-architectures — turns JTBD core-job research into grounded, superior user experiences via capabilities, skills, and workflows for product/UX work.
repos:
- path: ./packs/cognitive-architectures-maxmcp tools
Standalone MCP servers the packs spawn. Each resolves down a provider chain; Docker is the default because it's reproducible and sandboxed.
Heads up: the coordinates below — container images, release pages, registry ids — are the canonical targets for each tool. Publishing the actual artifacts is CI follow-up, so the install commands are marked pending publish until they go live. The shapes are real; the downloads aren't wired up yet.
Critical Path Method scheduling — earliest/latest starts, slack, the critical path, and bottlenecks — plus lock-aware cohort coordination for parallel work.
install optionsFailure Mode, Effects & Criticality Analysis — event-sourced FMECA fold, criticality matrix, and readiness scoring over an append-only session.
install optionsHuman-finished requirements elicitation — structured interview sessions and falsification-based discovery, resumable and append-only.
install optionsAn append-only scientific-process server — hypotheses, experiments, observations, conclusions, and verdicts, for evidence-driven work.
install optionsMultidimensional House of Quality (QFD) — cross-dimensional relationship matrices for structured trade-off and requirement analysis. (Available in the catalog; not yet wired into a pack's connections.)
install optionsBuild/log analysis — parse, grep, level counts, query, and summarize over raw log output (e.g. a full build to surface every issue).
install optionsMarkdown document administration — structured read/edit operations over Markdown files.
install optionscomposition
Load any number of packs, composed however fits your context — a proven baseline, a company policy layer, personal tweaks on top:
version: "1.0.0"
repos:
- path: ./packs/cognitive-architectures # the open baseline
- path: ~/repos/acme-policy # your company's gates — an example
- path: ~/repos/my-flows # personal tweaks — an example
# all namespaced + collision-checked at load; update any layer with git pullEvery layer loads under its own namespace. If two repos define the same id, praxec check fails loud rather than letting one silently win — a deliberate override has to be explicit, and it's audited. Like extending a base ESLint config or a Docker base image, but for what your workflows are allowed to do.
Note: acme-policy and my-flowsare illustrative names, not real packages. Repos are mix-and-match — organize them however fits your team.