a3s-code-core 9.0.0

A3S Code Core - Embeddable AI agent library with tool execution
Documentation
Summarise the conversation below into a compact context block. The summary will
replace the older messages so the agent can continue working with full awareness
of what has already happened.

Requirements:
- Keep the summary under 600 words.
- Preserve the **original user goal** verbatim in a "Goal" section. When a pinned
  goal is supplied below, copy it exactly into `## Goal` — do not paraphrase,
  shorten, or replace it with later assistant narration.
- List every file that was **read, written, or modified** with a one-line note on
  what changed.
- List every **shell command** that was run and its outcome (success / error / output
  snippet if important).
- Note any **errors or failures** encountered and how they were resolved (or that
  they are still open).
- Note the **current state**: what has been completed and what still needs to be done.
- Preserve the latest active user request, corrections, constraints, and immediate
  next action precisely.
- If the transcript already starts with a context summary, merge it into one
  updated cumulative summary. Do not nest or repeat old summaries.
- Do NOT include raw file contents or full command output — only summaries.

Output format:

## Goal
<original user request>

## Files Touched
- `path/to/file` — <what was done>

## Commands Run
- `command` → <outcome>

## Errors / Issues
- <error> → <resolution or status>

## Current State
<what is done> / <what remains>

---

Pinned original goal (copy verbatim under ## Goal; leave empty only if unknown):
{goal}

Conversation:
{conversation}