Context Router

context-router

A large codebase does not fit in a context window, and loading all of it makes an agent worse, not better — a model that has read everything has no signal about what matters. This skill replaces "read around until it feels familiar" with a traversal that terminates and that you can defend.

This skill owns the knowledge rule — the graph is the source of truth for structure and capability, loaded minimally — and the traversal algorithm that makes the rule executable.

The knowledge rule

The project keeps one LLM-maintained knowledge system at docs/graph/: Tier 1 routes, Tier 2 nodes own concise facts, Tier 3 leaf collections hold source-backed depth (libraries, provenance, product, architecture, APIs, data, prompts, evaluations, plans, runbooks, specs, decisions, tools). Never parallel doc systems.

• Load minimally, and declare it. Resolve the minimal node set from the router — entry nodes, their requires: closure, and only the composed depth the task names specifically — and declare what you loaded and deliberately skipped. Never bulk-read to get oriented; the graph is the orientation. The boundary you chose not to cross is part of the work's record, not a courtesy: a reader who cannot see it cannot tell an unread node from a read one. The algorithm below is the full form of this obligation; the delegation-boundary form is docs/graph/templates/prompts/graph-session-bootstrap.md.
• One home per fact. Every fact lives in exactly one node's owns:; everything else links. Duplicated facts rot asymmetrically. graph-lint.py enforces unique fact-keys, resolvable acyclic edges, and no version pin outside its owning library page. The code-side twin is one owner per concern: a cross-cutting behaviour — session and auth, policy emission, tolerant parsing, resource creation, a datastore — has exactly one owning component, and a second "convenience" mechanism for a concern that already has an owner is refused, because two mechanisms multiply precedence questions nobody can answer.
• Graph before code, ahead of memory. Memory of APIs and versions is unreliable; the graph is local and source-grounded. No wiki page for a library you're about to use → run ingest-library.
• The graph compounds. Record facts when code gains them, sharp edges when they bite, load_when: triggers when routing missed. Never fabricate a fact, version, or URL — write "not recorded".

Authoring and maintaining what this rule loads is skill.knowledge-graph (docs/graph/skills/knowledge-graph.md) — read its node contract once (the _schema.md the graph was built from); dependency leaves are skill.library-wiki. Prefer a configured Context7 / DeepWiki / llms.txt MCP server for fetching upstream content. If the project has no graph yet, build one via adopt-existing / knowledge-graph first; until then, fall back to reading the README and the plan-of-record, and say you did.

The algorithm

1. Classify the task in one sentence

Say what kind of work it is before deciding what to read. The four kinds route differently:

2. Resolve entry nodes

Open the graph's router index (Tier 1) and match the task against each node's load_when: triggers. Prefer the most specific match. A task naming a path resolves to that subsystem's node; a task naming a concept resolves to the node that owns it.

If nothing matches, you have found a gap in the graph. Say so, fall back to the root node, and note it for the graph's maintainer to fix.

A standard's match surfaces its standing exceptions. A deviation.* node (kind deviation, status: standing — the schema's "Node kinds") carries the departed-from standard's own name in its load_when, so a task that matches the standard's topic also matches every deliberate departure from it. Load it with the standard: a standing deviation is part of the answer, never a lapse to fix and never a decision to re-argue; its ends_when says when it stops applying.

Watch for aliased names across layers. When a subsystem answers to more than one name — a repo or folder name that differs from its product name, its package/artifact name, and its internal code name — a task that types one alias can silently fail to match a node keyed to another, and neither load_when matching nor a grep sees the miss. The Tier-1 router index must carry an explicit naming-divergence note listing the aliases for each such subsystem, so routing and search are not blind to any one of them. When you hit an unlisted alias, add it to that note in the same change (like sharpening a load_when trigger).

3. Take the closure

Load each entry node, then transitively load every node in its requires: list. That much you cannot be correct without. It is small by construction — if it is not, the graph is mis-modelled and should be fixed rather than worked around.

Then, from every loaded expertise node, take the composed children the task names specifically: descend into a child when the task uses, exactly, a term in that child's own vocabulary — its load_when: triggers plus its whole slug — that the parent does not already carry. So family words sitting on the parent descend nobody, and descent never folds a prefix the way the router's own entry matching does. A child you take becomes the parent for its own children, which is the whole of the recursion. composes: is a menu, not a closure: the specialisations the task is not about stay unread, and you say so (step 5).

A child you turn out to need but that descent did not reach is the same signal as a load_when: that should have matched and didn't. Load it, say you widened, and sharpen that child's triggers in the same change, so the next task routes there without you.

4. Do not take peers

peers: are the boundaries you are choosing not to cross. Load a peer only when the task explicitly crosses into it — and when you do, say why. The one exception is a trace: following a request or a message across subsystems is exactly what peers edges are for.

5. Declare before you work

Print the resolved set. This is not ceremony — it is the artifact that lets a reviewer catch a bad load before it becomes a bad change.

Task: add field <F> to <entity>                              [change]

LOAD (N nodes, ~T tokens)
  <entry node>       (entry)
  <required node>    (requires of <entry node>)
  <expertise node>   (requires of <required node>)
  <composed child>   (composed by <expertise node> on "<term>")

NOT LOADED (with the reason)
  <peer node>        peer of <entry> — owns X; not touched
  <peer node>        peer of <entry> — holds a copy of Y; cross only if
                     the change must reach it
  <sibling child>    composed by <expertise>; no task term specific to it

Tier-3 to open on demand
  <library page / spec / ADR> — if the detail is needed
  <expertise node>'s "depth" map names which leaf your identity needs

One NOT LOADED section, whatever kept a node out. A peer you chose not to cross and a specialisation the task never named are the same kind of record — the boundary, and the reason it held — and a set that lists only one of them hides the other.

Then, and only then, open source files — and only the ones the loaded nodes name.

6. Widen honestly, never silently

If mid-task you discover you need a node you did not load, load it and say so ("Widening: loading — the change is not local because …"). Silent widening is the failure this skill prevents. So is stubbornly working without a node you need in order to look disciplined. Both are worse than "I was wrong about the boundary."

Dry-run it

The graph router is executable and must run inside every spawned worker session — that requirement travels as the canonical block every delegation brief embeds (docs/graph/templates/prompts/graph-session-bootstrap.md: run --plan with the exact delegated task, load the closure, declare, return the output as route evidence); this skill owns the traversal algorithm above, not a second copy of that block. Dry-run your own hand-resolved set against the tool:

python3 <graph-tools>/graph-lint.py --plan "add a field to X"

If the two differ, one of you is wrong — usually a load_when: trigger needs sharpening, a cheap permanent fix.

--plan is a keyword heuristic, not an oracle. It ranks nodes by weighted term overlap; it does not reason about a request path or a false premise. Trust it for a single-subject change or question, and as a floor everywhere. But on four kinds of task, trust your own reasoning over its output:

• Traces — the right nodes are the hops on the path, which keyword overlap cannot infer.
• False-premise questions ("confirm we use X") — the correcting node may share no words with the wrong assumption; ask which node would own the truth.
• Policy questions — these have one owning node; --plan may pad the set. Prefer the single owner.
• Compound / multi-topic tasks — a task description that bundles several distinct topics dilutes each topic's distinctive terms below the keyword threshold, so the ranking can resolve to the wrong node and specialist set entirely, not merely a partial one. Probe each sub-topic separately, or explicitly discount the output for a task you know is compound.

Stopping rules

Stop loading when any of these is true:

• The closure is exhausted: requires: transitively, plus every composed child the task named specifically.
• You can state the change you are about to make and name the contract it must not break.
• The next node you would open is a peer the task does not cross into, or a composed child the task never named.

Do not stop merely because you have loaded "enough" files. The closure is the rule, not your comfort.

Cost discipline

• Within the loaded scope, retrieve progressively. The closure names the files; it does not license reading them whole. Indexes, headings, symbols, and diffs before regions; excerpts before full files; the complete source only when exactness demands it. Query in order of precision — exact identifier, exact phrase, constrained keyword, scoped filters, semantic search, broad exploration last — and let one authoritative source decide a question unless evidence conflicts or the consequence of error justifies corroboration (the retrieval posture: docs/graph/method/engineering-posture.md).
• A change task should load a handful of nodes. If it needs many, it is really several tasks; split it and say so.
• A trace may legitimately load many nodes along one path — but never a node off that path.
• Never load two sibling subsystem nodes "for comparison." If they are near-identical, what they share belongs in a shared node; read that instead. Needing a second sibling to infer a convention means the convention is missing from where it should live — add it there.

Anti-patterns

• Bulk-reading a subsystem to get oriented. The graph is the orientation. Confirming a path with ls/grep is fine; reading twenty files to build a mental model is the thing this skill stops.
• Loading the whole graph "to be safe." Full load is a summary that displaces the code you actually need — the most expensive way to know the least.
• Treating load_when as documentation. It is an index. When a task should have matched a node and didn't, fix the trigger in the same change.
• Skipping the declaration because the task is small. The declaration costs one paragraph and is the only record of what you did not read.

Reference files

• docs/graph/skills/knowledge-graph.md — builds and lints the graph.
• docs/graph/_schema.md — the node contract (the adopted copy beside the graph; docs/graph/templates/knowledge-graph/_schema.md is the pristine seed template, fast-forwarded on graft).
• docs/graph/templates/knowledge-graph/index.md — the router-index template.
• the kernel (AGENTS.md) — the context-budget rule this skill implements.