csa todo commands can fail with auto weave upgrade error before TODO lifecycle starts

[aider4ryder]

Summary

While running mktd planning flow inside a repository that already has a .csa/ directory, the workflow hit a CSA failure when it tried to inspect the existing TODO store.

The failing command was:

csa todo list

It failed with:

Error: auto weave upgrade failed after 3 attempts. Disable with [execution] auto_weave_upgrade = false

This blocks TODO lifecycle commands before planning can decide whether to create a new plan or replace an old one.

Environment

• csa 0.1.202 (v0.1.15-382-g3ad54bd)
• repo: RyderFreeman4Logos/mesh-proxy
• branch: feat/mktd-remaining-doc-implementation
• working tree contained an unrelated uncommitted change (weave.lock)
• csa config get execution.auto_weave_upgrade => true

Context

I invoked:

csa run --sa-mode true --tier tier-3-complex --no-daemon --timeout 3600 --skill mktd ...

The skill was planning implementation work from drafts/docs/00-06 and had already completed document/code reconnaissance. During the save / existing-plan inspection path, it attempted csa todo list and hit the auto-weave-upgrade error above.

Expected

• csa todo list should list TODO plans or return a TODO-specific error.
• Auto weave upgrade should not block TODO lifecycle commands in a way that makes planning unusable.
• If auto-upgrade fails, the error should clearly explain whether TODO commands can continue in degraded mode or why they cannot.

Actual

• csa todo list fails before the TODO lifecycle step can proceed.
• The caller cannot safely continue because mktd needs a real TODO save path/timestamp, not a fabricated one.

Why this matters

This breaks mktd end-to-end planning even when the repo is otherwise valid. In my case the planning workflow had enough RECON facts and was ready to draft the plan, but TODO lifecycle became unavailable due to the upgrade path.

Suggestion

• Consider making auto weave upgrade non-blocking for csa todo read/list/create operations.
• Or add a flag / config behavior that lets TODO lifecycle continue when auto-upgrade fails.
• At minimum, separate upgrade failures from TODO command failures more clearly.