Bind each lookup reply to its query, batch unique native callees, and fetch conservative literal string candidates while preserving Unicode, case, pagination, and raw source Evidence. Preserve explicit native search failures and describe unknown string representations.
Verified with Windows IDA Pro 9.3 and the native MCP plugin built from c9d15d98e8a143fc74f6e14037c48f63889b6d6c with a loopback listener. Production CLI/MCP parity and two-function switching passed on the public fixture; equivalent tasks passed on Windows whoami.exe. Windows IDA Free was not used. No personal paths or license registration information are included.
Validation: 182 focused tests, static checks, architecture guards, test discovery, and docs checks passed. Same-main measurements show warm task time 675 to 317 ms and estimated response tokens 16618 to 11509; cold-start, overall CPU, and memory improvement were not established. Detailed scope and measurements are in experiments/analysis-efficiency/ida-native-windows.md.
Record real Ghidra/MCP and Codex CLI verification of shipped ShareX native PE and FreeDOS DOS MZ entry views. Preserve original Evidence provenance, separate run-specific references from facts, and reject stale/colliding producer identities or unsupported model claims.
Sixteen focused experiment regression cases, original public-reply/transcript replays, and CI run 38273269072 pass for head 0667f7cb7c85f134311c7470ff4ee99ef9cc45f4. No production API/default change or general native speedup claim; final resource counters remain unknown. Continues #2016.
Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
Add opt-in retained inventory and WASM Evidence views with exact nested output contracts and a small static WASM descriptor. Preserve old complete CLI/MCP workflows, producer choices, original authenticated Evidence and cleanup ownership.
Validated with real public applications, genuine old Evidence migration, 123 focused assertions, 11 synchronized description checks and 42 installed-package WABT cases. Detailed generated experiment JSON stays outside the final tree; concise findings and reproduction inputs remain. Main-existing completion/package CI failures are separately attributed and left unchanged.
Refs #2016
Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>