[app-cli] `amplifier session list --format json` truncates session IDs, making the JSON output unusable for tooling

Observed on amplifier 2026.07.06-7ec5dcd: amplifier session list --format json -n 3 emits truncated IDs and a generic wrapper schema:

[
  {
    "name": "1326337c...",
    "enabled": true,
    "behaviors": ["1326337c..."],
    "config_summary": {
      "session_id": "1326337c...",
      "modified": "2026-07-08 16:01",
      "messages": "4"
    }
  }
]

Problem. JSON output exists for machines, but the 8-char+ellipsis IDs can't be fed back to amplifier session resume/run --resume reliably, and useful fields (working_dir, created, name/title, turn_count) aren't included. External tools are forced to read ~/.amplifier/projects/*/sessions/*/metadata.json directly instead.

Ask. In --format json mode, emit full session IDs and the session's metadata fields (session_id, created, working_dir, name, description, turn_count, parent linkage) with a schema shaped for sessions rather than the generic list wrapper.

Start by locating the Python implementation of amplifier session list and compare its --format json output with the metadata in ~/.amplifier/projects/*/sessions/*/metadata.json. Run amplifier session list --format json -n 3, then verify that full IDs and the requested session metadata are emitted in a session-specific schema usable by session resume and run --resume.