llms-txt

KINSHIP - llms-txt

LLM Read this

How to deduce gitub repos from path ?

• in example /workspace/repos/miadisabelle/mia-context7 the repo is miadisabelle/mia-context7
• Our jgwill/workspace (A Devops system), the pattern is as soon as we monitor with our Miadi-Webhooking system, the repo gets cloned in /workspace/repos/<owner>/<repo> and if a branch/PR is being developped, they are ofter checkout in that path.
• if cloned path is specified, the repo might also be in the workspace/repos but is worked on in that path for now, we plan to fully migrate to jgwill/workspace in the near future as an anthology to “Copilot Workspace” being deveopped in repo miadisabelle/mia-code-server that will influence conceptually and will provide sets of extensions to probably create jgwill/jgwill-code-server that the jgwill-workspace will use, this is said here so you know where we are going with this and that should point out that we might change how the jgwill/src are worked on, for now, it is a good place to be in the germination phase and as the vision gets assimilated and completed, migration toward a repo and being maintainted by jgwill/workspace is planned (examples of these path: /src/mcp-medicine-wheel*, /src/ceremony-session-observer/, /src/mcp-pde, and others that might be described in time… - example with /src/mcp-medicine-wheel* is the new jgwill/workspace#17 -package-> medicine-wheel-data-store + medicine-wheel-session-reader (jgwill/workspace#16) and you can observe there a pattern of not having a repo jgwill/medicine-wheel but the code being commited directly to the repo jgwill/workspace` simply because the OCAP isn’t clear on that (Who owns that “Medicine Wheel Development Suite” potentially applicable to two domains: ceremonial-technology-development (Decolonizing Sofware Development) and relational-science (Decolonizing Scientific Research).

Structural Tension Charting llms-structural-tension-charts.txt

repo: jgwill/smcraft

• Drafted guidance llms-stc-state-machine.md was created out of something Robert Fritz said that got imprinted in my mind: “The creative orientation is an Event-Driven Process” and Robert is not good with software architecture at all !! miadisabelle/forgewright#1 will be an interesting relation to apply that , learn from the relation to ceremony jgwill/src#403 and reflect that in our relational files.
• • Therefore, smcraft already has its first runtimes published at :https://www.npmjs.com/package/smcraft and https://pypi.org/project/smcraft/ and will be greatly influenced with its relationship with avadisabelle/coaia-narrative
• • (2026-08-14 — those two package names are now DEPRECATED. The runtimes ship as the @miadi/stateloom-* family on npm — engine, protocol, client, hub, react, canvas, cli, mcp, skills, web — and miadi-stateloom-engine on PyPI. The repo is still jgwill/smcraft. llms-stc-state-machine.md v2.0 carries the current names, the STC→MCP tool mapping and the PDE→chart→machine→rispec path.)

    repo: avadisabelle/coaia-narrative cloned in: /src/coaia-narrative

• create Structural Tension Charting with MCP and CLI tools.
• related guidance llms-managerial-moment-of-truth.md which should be available within coaia-narrative MCP tools

/src/coaia-visualizer/

• Web UI and network MCP to work with /src/coaia-narrative format

Digital Performance Review and Digital Decision Making Frawework

OLD repo git@github.com:miadisabelle/TandT.git cloned in /workspace/repos/miadisabelle/TandT and related guidances (llms-managerial-moment-of-truth.md AND llms-digital-decision-making.md)

• ALso was worked in /src/tandt/ which is where we extracted and created llms-digital-decision-making.md

• Question on TandT is : What in the place of the Evaluation model within the medicine-wheel ? how do we write on disk the various element of DESIGN

  EXECUTION that are being evaluated ? How does the evaluator process is triggered ? It is mandatory that human-in-the-loop used when elements of performance are being generated, evaluated because they touch human aspirations and AI Companions are at the level of Innovation, which is a secondary choice to create in relation to a primary choice that these criterias evaluate ?

new repo git@github.com:jgwill/veritas.git cloned in /workspace/repos/jgwill/veritas/ and related guidances (llms-managerial-moment-of-truth.md AND llms-digital-decision-making.md)

• Will offer a platform for models for the Performance Review type and the decision making type implied in the guidances (and related referenced internally).
• It become potentially an API / MCP that we can call to generate

/workspace/repos/jgwill/medicine-wheel/src/

• Bellow are subsections that the jgwill/veritas might feed on or the tool that consume the jgwill/veritas’s API would be built with (npmjs packages).

• jgwill/veritas and related guidances (llms-managerial-moment-of-truth.md AND llms-digital-decision-making.md) would have abstraction done to them for new components in either /workspace/repos/jgwill/medicine-wheel/src/ or /workspace/repos/avadisabelle/ava-langchainjs AND

  OR /workspace/repos/avadisabelle/ava-langgraphjs

• Medicine Wheel owns its technical reference at https://docs.medicine-wheel.jgwill.com/llms.txt and https://docs.medicine-wheel.jgwill.com/llms-full.txt. The llms-medicine-wheel-* files in this repository are permanent pointers only: do not copy package counts, versions, dependency ranges, tool inventories, or release data into them.
• TODO Go thru the path and upgrade this section in relation to what was created also in ./docs and recent packages for working with /src/mcp-medicine-wheel* futurely with data.

ceremony-protocol/

graph-viz/

narrative-engine/

ontology-core/

• I am suspecting that when we will want to generate models probably in relation to coaia-narrative with jgwill/veritas or whatever tools consuming that service, we will need some ontology to structure the model.
• • I have no idea of what will that look like

relational-query/

ui-components/

Extra-KinRepo

• repo/path that might influence how we distribute, access what is being created in here

/workspace/repos/miadisabelle/mia-context7

jgwill/etc-claude-code — cloned at /etc/claude-code on every host

This is where a guidance file stops being a document and starts being an instruction. /etc/claude-code/CLAUDE.md is organization-managed policy loaded into every agent session on the host, and its skills (rise-framework, pde-vessel-ceremony, …) read the llms-* files sitting beside it. A doc published here but absent there is written, not in effect.

The crossing is ./__SYNC_llms.sh, and it runs at git push:

./__SYNC_llms.sh --install-hook   # once per checkout: core.hooksPath = .githooks
./__SYNC_llms.sh --check          # what would travel, changing nothing
LLMS_SYNC_SKIP=1 git push         # publish without distributing

What travels is the union of __SYNC_manifest.txt (this repo’s canonical set — a fresh host bootstraps from it) and <target>/guidance.txt (host-local additions). The union is written back to the target’s guidance.txt, so that file stays an honest record of what is installed there rather than a wish list.

Two checks run on every pass, because a distribution that only copies cannot notice what it is failing to carry:

finding	meaning
REFERENCED-BUT-UNLISTED	the target’s CLAUDE.md/skills name an llms-* file that exists here and neither list carries — --adopt installs it, or add it to __SYNC_manifest.txt if every host should have it
DANGLING-REFERENCE	the target sends an agent to a file that exists neither here nor there — renamed, or never written

Exit codes follow the reconcile.py contract the ecosystem already uses: 0 clean or synced, 1 drift found, 2 could not look — the target is absent, unwritable, or not a git tree. That third code is the point. Before 2026-08-16 this script answered a missing target with exit 0, so the host that needed the sync most was the one reported as already done.

Related: jgwill/llms-txt#12 → ceremony jgwill/src#403.

This site is open source. Improve this page.