codex-agent-mem

Verification evidence

This directory contains reproducible, sanitized evidence for codex-agent-mem.

The current v1.0.0 public run was executed with Codex Desktop using GPT-5.4 in a Codex environment, reasoning effort xhigh. This evidence validates the Codex Desktop path; it is not a ChatGPT web/app connector validation.

All verification data is synthetic. The runner uses local SQLite files and local JSON/Markdown outputs only; it does not send memory, prompts, telemetry, or project data to external servers.

The v1.0.0 verification set covers:

• context compression and token savings;
• repeated-pack avoidance with known_pack_hash;
• lazy MCP initialization;
• minimal tool surface;
• read-only safety;
• response diet;
• local telemetry;
• closure-control checks;
• a sub-agent handoff scenario.

Start here:

• v1.0.0/RESULTS.md
• v1.0.0/results.json
• v1.0.0/METHODOLOGY.md
• v1.0.0/checksums_sha256.txt

This site is open source. Improve this page.