Report

halo_report and halo trace report return the same Markdown. It has four parts.

1. Header. The number of events and sessions, the time span, and the level the trace supports.
2. Channels. One row per channel. S→A is marked basic because every agent system has it.
3. Evidence for self-evolution. One row per evidence step, then one connected chain if the trace has one: an action, an outcome about it, a revision that used the outcome, and a later action that used the revision.
4. Levels. One row per level from L0 to L5, with the events that meet each condition, or what the level still needs. A last line names what the next level needs.

Cells

Value	Meaning
observed	the trace contains links that meet the condition
not observed	the trace contains no such links, which does not mean the agent lacks the flow

Evidence notation

Notation	Meaning
e3→e5	event e5 used event e3
e5⇠e7→e9	outcome e7 is about action e5, and e9 used e7
e9	the event itself

At most three examples are listed per cell.

Example

The report of the dot in ChatGPT dots, abridged:

**The trace supports L0 (Session local).**

| Channel | In this trace | Evidence |
|---|---|---|
| S→A (basic) | observed | e1→e4, e2→e4, e1→e5 |
| S→M | observed | e2→e7, e6→e7 |
| M→A | observed | e3→e4, e3→e5 |
| A→S | observed | e5→e6 |
| M→S | not observed |  |
| A→M | observed | e4→e7, e5→e7 |

| Step | In this trace | Evidence |
|---|---|---|
| attribution | not observed | outcomes name their action, but no later action or revision used them |

To support L1, the trace needs something kept in memory and used in a later session.