route
Skill4 starsChanged 48 days agoMITView on GitHub
Route a requirement to the project(s) that own it via the project registry, read each project's runtime tier, then delegate a tier-appropriate brief — splitting into parallel briefs when the requirement spans more than one project
17 other projects use a version of this.
Adopted by
Most recently taken up by
- jerseycheese/agent-skillstheir own version, first seen Oct 28, 2026
- Hugovdd/skillstheir own version, first seen Oct 25, 2026
- rehantariq121/claude-skills-publictheir own version, first seen Oct 20, 2026
- RumbleJack56/skillstheir own version, first seen Oct 20, 2026
- 0xcvaliente/claude-code-orchestrationtheir own version, first seen Oct 17, 2026
- anton-dergunov/agentic-org-plannertheir own version, first seen Oct 13, 2026
Used with
Projects with it also use prototype.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
GoogleSkillprototype
Build a throwaway prototype to answer a design question. Use when the user wants to sanity-check whether a state model or logic feels right, or explore what a UI should look like.
Also in 3 projects that use this
Other versions17 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 17
- ypollak2/llm-router★ 123same name
- mehrad-dm/mastermind★ 23same name
- ItsssssJack/seven-skills★ 19same name
- purpleeddy/cladding★ 16same name
- dawidtomaszewskipl/route-skill★ 4same name
- gabriellst/codm★ 4same name
- jerseycheese/agent-skills★ 4same name
- 0xcvaliente/claude-code-orchestration★ 2same name
- Deepanshu418/claude-code-elysia★ 2same name
- RumbleJack56/skills★ 2same name
7 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository12 other files it gives its agents.
- code-reviewSkill
- context-healthSkill
- curateSkill
- demand-pipelineSkill
- find-skillsSkill
- learnSkill
- onboard-projectSkill
- open-prSkill
- planSkill
- portfolio-triageSkill
- releaseSkill
- verifySkill
The file
We split this skill into the rules it gives an agent
3 rules. 0 are common, 1 is distinctive, and 2 are very specific to this project.
0
Common
Most of the 16 edited versions we read keep these.
1
Distinctive
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 47
Record before delegating: Write the task into the owning project's board (docs/tasks/<project>.md, per the durable-task-state discipline) …
- Very specific: line 55
Track to completion: When a coworker ACKs, update that project's board (move the item to *Recently shipped* — with the commit SHA for a …
- Distinctive: line 62
Every requirement has exactly one place it belongs.
“Common” is measured: kept by at least 6 in 10 of the 16 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Turns "here's a thing that needs doing" into "here's who does it and the brief they're working
from." The manager's entry point for a single inbound requirement (contrast with portfolio-triage,
which surfaces work proactively across everything).
Arguments: route <requirement description or task-board item>
(Optional prefix: route --project <name> <requirement> to skip lookup when the owner is already known.)
Steps
- Read the registry. Parse the live registry
docs/portfolio-registry.md(created fromtemplates/project-registry.md) — ONLY the rows between its<!-- projects:start -->/<!-- projects:end -->markers. That block is the routing table (project · tier · responsibilities · keywords). IGNORE every other table in the file — the tier→policy table uses the same pipe/backtick shape and is NOT a project list. Seedocs/portfolio.md§3. Do not guess an owner from the requirement text alone; match it against the registry's stated responsibilities. - Identify the owning project(s).
- Single owner — the common case. Confirm the match against the registry entry's stated responsibilities, not just a keyword hit (a keyword can be ambiguous across two projects).
- Spans multiple projects — split into one independent requirement per project. Each split must be self-contained: a sub-agent executing it should not need the other projects' context to do its part. Note the dependency order if one project's output blocks another's (e.g. a shared schema change must land before the consumer).
- No match — the requirement doesn't map to any registered project. Do not force it onto the
closest-sounding one; either register a new project first (see
docs/portfolio.mdfor the onboarding steps) or hand it back for scoping.
- Read the owning project's runtime tier off the registry's Tier column (
docs/portfolio.md§2) —personalorwork. This decides the handling the brief must carry, not just which repo to touch:personal→ the brief says: commit direct tomain, push when done, no branch, no PR.work→ the brief says: create a feature branch, open a PR, and stop there — do not merge without the review gate. Note the worktree the coworker should use for isolation. For a cross-project split, read each project's tier independently — a single requirement can legitimately fan out into onepersonal-tier brief and onework-tier brief with two different handlings.
- Enrich each brief before dispatch. A bare requirement sentence is not a brief. Add: the
objective, the acceptance criteria (what "done" looks like, concretely), the tier's handling from
step 3, and any constraint the owning project's
agents/<project>.mdcalls out (must-NOT-do scope, conventions). A coworker should never have to re-derive intent, or the right-sized workflow, from a one-liner. - Record before delegating. Write the task into the owning project's board
(
docs/tasks/<project>.md, per the durable-task-state discipline) as in-progress, so it survives a/cleareven if the delegation itself is still running. - Delegate. Dispatch each project's brief (addressing and submitting the dispatch is your own execution setup's job — out of scope for this repo). Multiple projects → multiple parallel dispatches in the same pass — don't serialize independent work. Same-project multiple tasks may be merged into one brief or sequenced by dependency.
- Track to completion. When a coworker ACKs, update that project's board (move the item to
Recently shipped — with the commit SHA for a
personal-tier project, or the PR link for awork-tier one — or the next queued step) before considering the route closed. A route that dispatches but never confirms completion isn't done — it's dangling.
Why this exists
Every requirement has exactly one place it belongs; guessing costs a wrong delegation and a redo.
route makes "who owns this" a lookup against a written registry instead of a judgment call repeated
from scratch every time, and makes "does this span projects" an explicit branch instead of an
implicit assumption that everything is one project's problem. Reading the tier before dispatch is
what stops a route-generated brief from imposing the wrong workflow — PR ceremony on a solo
project, or a direct-to-main commit on a shared one — which is the single most common way a
portfolio-wide routing skill can silently do damage.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

