| name | aoa-session-memory-stress-pass |
| description | Use for `.aoa` large-session checks, logical compaction boundary validation, bounded archive inspection, and `stress-pass` diagnostics. |
| license | Apache-2.0 |
| metadata | {"aoa_scope":"session-memory","aoa_invocation_mode":"manual"} |
aoa-session-memory-stress-pass
Use when a large session archive needs bounded validation without flooding the
active context.
Trigger Boundary
- The user asks for a large archive check.
- Segment counts, compaction counts, or marker counts look suspicious.
show --full would be too large.
- A prior
raw_unavailable or rebuild may have damaged confidence in an
archive.
Procedure
- Read
AGENTS.md, DESIGN.md, DESIGN.AGENTS.md, and PIPELINE.md.
- Resolve the target session from
session-registry.json; default to
latest only when the user did not name a session.
- Inspect
session.manifest.json and segment indexes before opening bulk
Markdown or raw JSONL.
- Run:
python3 scripts/aoa_session_memory.py stress-pass <session> \
--aoa-root <aoa-root> \
--compactions 100 \
--write
- Use
--write only when a durable diagnostic was explicitly requested.
That write is limited to generated diagnostics and never authorizes archive
or raw-evidence mutation. Keep stdout bounded and read the written
JSON/Markdown artifact when detail is needed.
- Distinguish logical
compaction_boundary_count from raw marker count.
Verification
ok=true in the stress-pass payload.
- The selected segment span is coherent.
- No marker-only microsegments were introduced.
- Written diagnostics exist under the session
diagnostics/ directory.
Stop Line
Do not use --full unless the user explicitly needs complete JSON on stdout.