aayudha docs Built from de79988 (2026-10-26)

chart: design the data or AI behind a change

A dashboard nobody decides from, or an AI answer nobody can trace, is expensive to build and hard to undo. chart interviews you before anyone plans the change, then posts one chart note on the issue: the decision it serves, the data or evidence that answers it, and how to prove it right.

Field Value
In plain words Before a data or AI change is planned, works out with you which decisions it serves, which data or evidence answers them, and how to prove it right.
Journey stage Break it down
Accountable seat Data & AI Architect (Shuka)
Does the work Shuka, the Accountable seat's twin, in the spec model lane, with the seat's paired human when it has one
Consulted Code Architect (Vishwa), Revenue Architect (Kuber)
Informed AI Native Engineer (Nakul), Quality Architect (Yama)
Layers it can touch L05 Intelligence, L07 Operational data / state, L08 Analytics / reporting
Who reasons your agent's own model, inside the skill
Writes as the bot identity (a GitHub App); without one, your identity, tagged as human fallback
Where it runs the aayu command, a skill in your agent, an MCP tool (read-only unless [mcp] expose_writes is on)
The skill your agent reads skills/chart/SKILL.md

When to use it

An AI feature that also reports numbers gets two notes, one of each kind. Ready to lay out the build? Use plan, which reads the note.

Run it

Ask your agent with the issue's number and the kind. It asks two to four questions a round, riskiest first, each with a default, then reads the design back to you in three lines before it posts.

Use the chart skill on Feature 42 with kind data: work out with me which decision the dashboard serves before
anyone plans it.
aayu chart <issue>
Input Required What it does
<issue> yes Feature, Initiative, Story, or Bug whose data or AI design to chart.
--kind yes data: decisions, metrics with grain and window, sources, platform and its trigger, retention and deletion; ai: simpler options, evidence and permissions, abstention, and an evaluation set. One of: data, ai.
--draft no Markdown chart note the host's model wrote; chart checks every part, the stage's artefact, and every citation. Without it, chart prints the grounding and the outline.
--create no Post the checked note on the issue as one comment per kind, replaced in place on a rerun (through the actions log); needs --draft.

The final step is aayu chart 42 --kind data --draft note.md --create.

What you get back

One comment on the issue per kind, replaced in place when you run it again. A data note covers the decision, the metrics with their grain and window, the smallest design that serves today's decision and the trigger for the next, privacy and retention, and the checks. An AI note covers the question, simpler options, evidence and permissions, when to abstain, and an evaluation set with adversarial cases. Its length follows your team stage: half a page at a startup.

You should see a comment on the issue whose title names the note, such as # Data sketch: … at a startup. plan then builds the story's checkpoints and tests from it.

Gates

It checks these before it acts, and stops when one fails:

Before posting, --draft checks every part of the note and every citation. A claim the grounding does not show is refused unless it is marked Assumption: or Unverified:.

What only a person does

The registry reserves no step of this verb for a person.

chart writes no code and touches no data. A design choice that changes what data is kept or who may read it is yours: the agent offers a Decision draft, and you accept it with decide.

What's next