| name | tableau-route |
| description | Router for the tableau-dashboard-plugin workflow. Reads the project's STATE.md and reports which single skill the analyst should run next, honoring the ordering rule. Use when the user asks "what's next", "where am I in the workflow", or "which tableau skill do I run now". It only recommends — it never runs another skill itself. |
| disable-model-invocation | true |
| allowed-tools | Read, Bash(python *), Bash(python3 *) |
tableau-route
A thin router over the file-based workflow contract. It reads STATE.md in the
analyst's project and tells them which one skill to run next — nothing more.
This skill is a router only. It NEVER invokes another skill inline. Running
the next skill here would share this conversation's context and defeat the
entire point of the plugin (each skill must run in its own fresh conversation).
Your job ends at recommending the next skill and telling the analyst to run
it themselves.
How to run
-
Determine the analyst's project directory (the directory that contains, or
should contain, STATE.md). Default to the current working directory.
-
Run the router script and capture its output:
python "${CLAUDE_SKILL_DIR}/scripts/route.py" "<project-dir>"
(Use python3 if python is unavailable.) The script reads STATE.md,
applies the ordering rule from CONTRACT.md, and prints the recommended next
skill plus a one-line reason. It is pure and read-only — it does not modify
any files.
-
Relay the script's recommendation to the analyst, then stop. Tell them to
open a fresh conversation and invoke the recommended skill themselves (e.g.
"Run tableau-data next"). Do not call the recommended skill yourself.
Interpreting the result
The script reports one of four situations:
- fresh project (no
STATE.md) → recommends tableau-init to scaffold the
project.
- a step is ready → recommends that step's skill (the first step whose status
is
pending or stale, with its required inputs satisfied).
- blocked → a required upstream artifact is unresolved or missing on disk;
the script points at the upstream skill to run first.
- done → every step is
approved or skipped; point the analyst at their
deliverables under mock-version/v_N/.
If STATE.md looks malformed or the recommendation is surprising, show the
analyst the relevant part of STATE.md and refer them to CONTRACT.md (the
canonical schema) rather than guessing.
The full schema and the ordering / staleness / versioning rules live in
CONTRACT.md at the repo root. This skill deliberately restates only what it
needs; route.py is the executable mirror of that contract.
