fix: omit non-terminal runtime promotion writes

[haasonsaas]

Summary

• omit dry-run terminal promotion writes when the final local ledger state is still running, waiting, or blocked
• keep governance decisions from being projected as active wait operations
• add focused ledger tests for non-terminal and governance promotion behavior

Verification

• node ./scripts/run-vitest.js --run test/cli/run-command.test.ts test/server/agent-trajectory-replay.test.ts test/server/agent-trajectory-validation.test.ts
• bunx tsc -p tsconfig.build.json --noEmit

Follow-up to #419 and mirror for evalops/maestro-internal#1974.

[cursor]

PR Summary

Medium Risk
Changes how dry-run promotion decides whether to emit complete_run/fail_run and how blocked/governance events map into Platform-shaped operations, which can affect run finalization semantics. Coverage is improved with new focused tests, but edge cases in terminal-entry selection could still alter promotion outputs.

Overview
Tightens dry-run AgentRuntime promotion planning to avoid emitting terminal complete_run/fail_run operations unless a substantive terminal ledger entry exists (skipping passive kinds like run/wait/governance and adding explicit warnings when omitted).

Adjusts promotion shaping by not projecting governance events as active waits (waitType now only for wait) and by mapping blocked ledger states to a valid step state for record_run_step (coerced to failed). Adds targeted run-command ledger tests covering non-terminal finals, passive terminal entries, governance behavior, and blocked-state mapping.

^{Reviewed by Cursor Bugbot for commit 913d84c. Bugbot is set up for automated code reviews on this repo. Configure here.}

[github-actions]

This PR changes mirrored Maestro source files in the public repo, but it does not link the matching private source-of-truth PR.

Add one of these to the PR body, then re-run the check:

• https://github.com/evalops/maestro-internal/pull/<number>
• evalops/maestro-internal#<number>
• maestro-internal#<number>

Mirrored files touched:

• src/server/agent-runtime-ledger.ts
• test/cli/run-command.test.ts

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 95f4fd6d4b

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 913d84c2e9

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[chatgpt-codex-connector]

Preserve wait_run emission for pending governance entries

This change removes waitType for all governance entries, so a policy.decision that maps to status: "pending" (which becomes ledger state waiting) will no longer generate a wait_run operation. That drops the only operation carrying approval-wait semantics (waitType) for those trajectories, so downstream runtime promotion can no longer represent the run as actively waiting for approval even though the ledger step/work item is waiting.

Useful? React with 👍 / 👎.