Navigation
Open-source agent skills for building scope-safe Markdown knowledge vaults with deterministic validation and research workflows.
npx -y skills add pariyar07/ariadne --skill navigationAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 18 stars18 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Design or modify Markdown knowledge-vault navigation: folder hubs, agent navigation, routing matrices, workstream graphs, optional Bases/view layers, templates, health-check rules, and long-term traversability.
SKILL.md
7.5 KB, as published. Nobody here has run it
Ariadne Navigation
Use this skill when changing the durable structure of a Markdown knowledge vault.
Core Principle
Optimize for reliable human and agent traversal, not for a beautiful graph view.
Markdown notes remain the source of truth. Bases, canvases, dashboards, and reports are view layers over the vault.
Start
- Read root instructions and every applicable ancestor scope instruction through the target.
- Read relevant indexes, agent navigation, navigation standards, routing matrices, and knowledge-processing architecture.
- In a multi-scope vault, require a current-turn named or confirmed target before writing.
- State an explicit
allowed_write_set. Name every parent hub, routing matrix, root Base formula, or cross-scope file separately; minimal wiring is not implicit authorization.
Navigation Layers
Maintain these layers when useful:
- Global entry:
00 Index.mdand agent instructions. - Agent entry:
Agent/00 Agent Navigation.md. - Folder hubs:
00 ... Index.mdnotes inside durable folders. - Thread hubs: synthesis or debate notes for recurring arguments.
- View layer:
Bases/00 Bases Index.md,.basefiles, canvases, dashboards. - Note-level links: meaningful links to sources, concepts, entities, relationships, questions, and decisions.
Workstream Graph Protocol
Do not pre-create graph categories because they might be useful later.
Create or promote a workstream graph only when a recurring use case needs its own traversal path. Examples can include evidence trails, customer discovery, engineering decisions, GTM experiments, fundraising, life systems, or any other durable domain.
When promoting a workstream:
- Choose or create the folder.
- Create a
00 ... Index.mdhub. - Add a routing row to
Agent/Task Routing Matrix.md. - Link from
Agent/00 Agent Navigation.mdif it becomes a primary route. - Add templates only if the structure repeats.
- Add a Base only if metadata inspection helps.
- Link the Base from
Bases/00 Bases Index.md. - Update
00 Index.mdonly if strategically important. - Add health-check coverage if the workstream creates recurring maintenance.
Recursive Scope Protocol
A navigation scope can exist at any depth: the vault root, a top-level workstream, a child workflow, or a deep child project. Promote a child scope only when a recurring durable route needs local traversal; otherwise keep it as a normal folder under the parent scope.
Topology ownership is strict: sync_scope_topology.js owns descriptors, generated files, generated blocks, the scope map and Bases/Scope Registry.base, routing checkpoints, redirects, and parent/child topology. Use ariadne:scope for any lifecycle or topology change. Non-registry root Base formulas are report-only: they remain user-authored and may be updated outside generated blocks only through a separately confirmed explicit allowed_write_set.
Navigation may edit only user extension areas outside generated blocks: prose around marker-managed checkpoints, user-authored hub sections, semantic links, and other explicitly confirmed content. Never edit inside generated blocks or reproduce generated wiring elsewhere.
When improving a child scope's user-authored navigation:
- Identify the parent scope and its hub.
- Create or update the child hub, usually
00 ... Index.md. - Preserve engine-owned parent/child links and routing checkpoints.
- Add only semantic user-authored links outside generated regions.
- Add a local
AGENTS.mdonly for local deltas. - Add local templates or local Bases only when the repeated workflow justifies them.
- Add health checks for hub links, scope queues, raw-source compilation, and Base coverage.
Warn when you see:
- child
AGENTS.mdfiles repeating parent or root policy instead of local deltas - child local Bases without scope filters
- parent hub missing child link
- child hub missing parent link
- ambiguous cross-scope wikilinks that should point through the nearest useful hub, relationship note, or synthesis
Bloat And Drift
When navigating or restructuring a vault, proactively call out:
00 Index.mdacting like a complete table of contents instead of a strategic mapAgent/00 Agent Navigation.mdcarrying detailed note lists instead of routes- folders with many durable notes but no
00 ... Index.md - hubs that are too long to scan and need sub-hubs or a Base
- recurring workstreams missing from the task-routing matrix
- Bases not linked from
Bases/00 Bases Index.md - raw/inbox/output/question buildup without a compilation route
- repeated need for user guidance because global instructions are too generic for a folder
If the current task is structural, fix the issue directly when safe. Otherwise, tell the user briefly and propose a focused maintenance pass.
If a proposed fix changes scope topology, do not fix it directly: route it through sync_scope_topology.js via ariadne:scope.
Folder Index Rules
Use:
Top-level index = one per durable workstream
Nested index = only when a subfolder becomes a recurring workflow
Thread hub = when a topic becomes an ongoing argument or synthesis
Base = when dynamic filtering or status inspection helps
Local AGENTS.md Pattern
When a specialized folder needs local rules, add or propose a folder-level AGENTS.md.
Use local AGENTS.md files for:
- folder purpose and boundaries
- local note types and required frontmatter
- local workflow
- templates and Bases to update
- done criteria
- local bloat/drift signals
Do not create local AGENTS.md files just to repeat global rules. A good trigger is repeated user guidance for the same folder-specific behavior.
For AGENTS.md, CLAUDE.md, or GEMINI.md files in an external code repository or ordinary folder, use ariadne:workspace-instructions. This navigation skill owns vault-local hubs, routing, and folder rules; workspace instruction files own the bridge back to registered Ariadne context.
Base Layer
Base files should be linked from Bases/00 Bases Index.md with wikilinks, for example [[Bases/Decisions.base]].
Validation should resolve both Markdown notes and .base files.
Guardrails
- Avoid decorative links.
- Avoid creating wrapper notes only to make non-Markdown files appear pretty in graph view.
- Keep
00 Index.mdselective. - Link operational files through operating hubs; link semantic notes through semantic context.
- Read raw sources only when compiled notes are insufficient.
Validation
Before finishing:
- run or describe link validation
- confirm new hubs are linked from the right parent hubs
- confirm Bases are linked from
Bases/00 Bases Index.md - confirm task routing points future agents to the smallest useful context set
- confirm no existing workflow was silently replaced
Related Skills
- Use
ariadne:vaultfor new vault bootstrap. - Use
ariadne:workspace-instructionsfor workspace instruction files outside the vault. - Use
ariadne:workstream-trackingfor Kanban work boards and Dataview dashboards. - Use
ariadne:maintenancefor health checks. - Use
ariadne:research-pipelinefor domain research pipeline setup. - Use
ariadne:research-ingest,ariadne:knowledge-capture, andariadne:research-synthesisfor content flow. Research topology inside a confirmed boundary belongs toariadne:research-pipeline; durable cross-scope navigation remains here.