Sculptured docs

Context and communication

Inspect captured context, tool results, messages and usage, with the source of each claim.

Sculptured's progress views summarize captured messages, tool calls, file changes and checks. Use them to understand a run, then follow individual entries back to their recorded evidence when you need more detail.

ViewWhereAnswers
Visual progressAgent page, Activity tabWhat did this agent build, run and check?
ContextAgent page, Activity tab, inside Visual progressWhat went into this agent's context?
CommunicationsRun board, MoreWho talked to whom, and how much?
ConversationRun board, MoreWhat happened in the team, in order?
Token useRun board, MoreHow many tokens did each agent report?

Read the evidence labels

Each label identifies how Sculptured obtained the information. It does not assign a general confidence score.

LabelMeaningExample
CapturedRecorded word for word as it happenedA tool's input and result
ObservedSeen by the computer running the agentThe process started or ended
Agent-reportedThe agent said so"Done", or findings from report
HeuristicSculptured's inference"Stalled", a stale agent, an estimate
Verified checkA command Sculptured ran itself, with its real exit statusYour test command passed

Visual progress

Open an agent and choose Visual progress.

Pick any item to jump to that moment in the Transcript. The Transcript then offers a way back to Visual progress.

An agent's context

The Context section of Visual progress shows what Sculptured captured going into the agent, by category:

CategoryWhat it holds
Team & job instructionsThe team briefing and the agent's Job
Your messagesWhat you sent it
Assignments & handoffsWork handed to it by other agents or a workflow
Tool inputsWhat it asked tools to do
Tool resultsWhat tools gave back
Agent repliesWhat it wrote

Each category shows records, characters, and ≈ tokens, estimated as characters ÷ 4. Open a category with Inspect captured content, or list the files and links mentioned with Inspect referenced paths and URLs.

Above the categories is the last reported context usage, labelled "CLI-reported usage" or "CLI-derived estimate", with its time. If the CLI has not reported yet, it says so.

Team communication

Open the project's Results tab, then Communications.

PartWhat it shows
FiltersOne participant or Everyone, the window (Loaded history or Last hour), and direction
TotalsMessages, read and failed counts, captured payload size
Who talks to whomThe routes messages actually took, as a picture
Message trafficMessages over time
Reported model usageTokens the CLIs reported
RoutesEach sender and receiver pair, with messages, payload, and read or failed
MessagesEach message, with Open receiving transcript

An arrow on the Design board describes the intended route. This view shows captured messages from the selected history window. An arrow with no recorded traffic has no handoff evidence in that window.

The team conversation

Results › Team conversation puts the team's updates, messages, handoffs, questions and reported results on one timeline. Filter it to All communication, Updates, Messages & handoffs, Questions or Reported results, or to one agent. Each row opens the agent's conversation, the work evidence, or the captured event behind it.

Token use

Results › Usage charts the tokens each agent reported over time: output, cached input read, and context now, with each agent's last report. Usage arrives when the CLI reports it, often at the end of a turn, so the chart can lag the work.

Raw events

The Events tab lists the recorded events underlying the progress views, in order and with their sources. Use it to investigate a summary or status that needs closer inspection.

Next steps