chore: sync public mirror from internal

[haasonsaas]

Summary

• sync the sanitized public tree from evalops/maestro-internal
• keep evalops/maestro as a generated public mirror of the private source of truth
• preserve public-owned CI and trusted-publishing workflows from the public checkout
• internal source SHA: b8348e9436d58002960281544e85685c9c7e021e
• last generated public sync base: 8a55f9467403be9dbe97126916a697e64e733c4a
• previewed public-tree drift: 3 file(s) to copy/update and 1 stale file(s) to delete
• public-only commits since last generated sync: 0

Source-of-truth status

Public Mirror Drift Audit

• package: @evalops/maestro
• private source: https://github.com/evalops/maestro-internal@main (b8348e9436d5)
• public projection: https://github.com/evalops/maestro@main (8a55f9467403)
• files to copy or update: 3
• stale files to delete: 1
• result: drift detected
• invariant: public_projection_has_drift

Sample Changed Paths

• copy/update src/platform/agent-runtime-client.ts
• copy/update src/platform/cerebro-facts-client.ts
• copy/update test/platform/agent-runtime-client.test.ts
• delete src/utils/abort.ts

Guidance

Let internal main generate and merge the public sync PR before relying on public main.

Drift sample

• copy/update src/platform/agent-runtime-client.ts
• copy/update src/platform/cerebro-facts-client.ts
• copy/update test/platform/agent-runtime-client.test.ts
• delete src/utils/abort.ts

Public-only commits since last generated sync

• none detected since last generated sync

Validation

• generated by the sync-public-release-mirror workflow in public-tree mode

Supersedes

• updates existing generated public sync PR chore: sync public mirror from internal #186 to internal source b8348e9436d58002960281544e85685c9c7e021e

[cursor]

PR Summary

Medium Risk
Moderate risk: changes the shape of MaestroFactsContext and adds new Cerebro MapThing calls (with partial-failure handling), which could affect downstream consumers and runtime trigger payload size/behavior.

Overview
Maestro session trigger enrichment now gathers a richer Cerebro facts context: Search responses can include links, and each found thing is additionally expanded via a new MapThing call to collect related things, links, paths, and extra evidence (deduped), while tolerating individual map failures (except aborts).

MaestroFactsContext is extended to include linkIds, eventIds, links, paths, watermarks, and an expanded summary (link/path/watermark counts), and tests are updated to validate the new payload contents, request ordering, and cancellation propagation. The shared isAbortError helper is inlined where needed and src/utils/abort.ts is removed.

^{Reviewed by Cursor Bugbot for commit d1f5a60. Bugbot is set up for automated code reviews on this repo. Configure here.}

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: bbd56c5990

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: 89044f66ba

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]

Make MapThing fetches resilient to per-thing failures

This Promise.all makes gatherMaestroSessionFactsContext fail fast if any single MapThing call errors (e.g., one timeout/404 among multiple thing IDs). In recordMaestroSessionRuntimeTrigger, that rejection is caught and factsContext is dropped entirely, so the runtime trigger loses all facts (including already-fetched Search/GetThing data) instead of degrading partially. This introduces a new all-or-nothing failure point for facts enrichment.

Useful? React with 👍 / 👎.