docs: update slice-dag and orchestrator-cli for #2399/#2403/#2410 [doc-updater]

[james-in-a-box]

Update documentation to reflect changes from a505cfa (PR #2402):

• Branch namespace (Slice integration branches collide with pipeline branch in git ref namespace ('directory file conflict') #2399): The pipeline tip is now pushed to egg/<id>/work so slice integration branches (egg/<id>/slice-N) can coexist as siblings. The get_worktree_branch table and get_slice_integration_branch description in slice-dag.md were showing egg/issue-N as the pipeline branch shape; updated to egg/issue-N/work with an explanation of the git "directory file conflict" constraint that drove the change.

• Slice-scoped signal routing (Slice agents spawned with invalid EGG_PIPELINE_ID='<id>/slice-N' and no worktree mount #2403): EGG_SLICE_ID was previously described as an "advisory hint" in the slice-dag doc. BRC handlers now read it and explicitly attach slice_id to every CONSENSUS_* payload so the orchestrator routes to the per-slice tracker. Updated the description accordingly.

• restart_agent with slice_id (Plumb slice_id through restart_agent_job and detect_uncommitted_changes (#2403 follow-up) #2410): Removed restart_agent with slice_id from the "Out of scope (Independent implement phases #2137)" list in slice-dag.md — the REST endpoint now accepts ?slice_id=slice-N. Added a slice-scoped curl example to the agent restart note in orchestrator-cli.md.

Triggered by: #2402

Test Plan

• Automated: no code changes; documentation only.
• Manual: verify the updated branch shape and EGG_SLICE_ID descriptions match the implementation in orchestrator/routes/pipelines.py (_ensure_pipeline_work_ref) and sandbox/egg_agent_tools/handlers/brc.py (_maybe_attach_slice_id).

Authored-by: egg

[egg-reviewer]

No agent-mode design concerns — documentation-only PR updating descriptions of branch namespace, slice-scoped signal routing, and a restart_agent curl example. No prompts, agent surface, or control-flow changes to assess.

— Authored by egg

[egg-reviewer]

Approve — documentation-only update; all three claims verified against the implementation

This PR updates slice-dag.md and orchestrator-cli.md to reflect three landed changes (#2399, #2403, #2410). I verified each claim against the current code:

1. egg/<id>/work branch namespace (#2399)

_ensure_pipeline_work_ref in orchestrator/routes/pipelines.py:682-726 performs the documented normalisation (egg/<id> → egg/<id>/work, idempotent on egg/<id>/work, structural ≥2-slash check so egg/work deepens to egg/work/work). The "directory file conflict" rationale added to the get_slice_integration_branch paragraph matches the constraint the helper guards against.

2. slice_id attachment to CONSENSUS_* payloads (#2403)

sandbox/egg_agent_tools/handlers/brc.py:26-41 defines _maybe_attach_slice_id, which reads slice_id from the request or falls back to get_slice_id() (the EGG_SLICE_ID reader) and validates against slice-<N> before attaching. It is called from all 5 BRC signal handlers (lines 444, 529, 586, 634, 754). The previous slice-dag.md text describing EGG_SLICE_ID as "an advisory hint" is genuinely stale; the new wording is accurate. The cross-reference to peer_consensus._tracker_key is correct — orchestrator/peer_consensus.py:1834 composes {pipeline_id}/{slice_id} when slice_id is supplied.

3. restart_agent with slice_id (#2410)

orchestrator/routes/pipelines.py:2262-2338 exposes the documented surface — slice_id accepted via query string with body fallback (query param wins, both validated through extract_slice_id). The "Out of scope" list edit in slice-dag.md correctly removes restart_agent with slice_id and notes the landing PRs. The new curl example in orchestrator-cli.md:211 is well-formed (URL is quoted, which is needed for the ? query separator).

Other observations

• No new code; no test plan beyond manual doc/code cross-check is appropriate here.
• Heading levels, table formatting, and link targets unchanged outside the edited regions.
• The _run_concurrent_phase paragraph's mention that _handle_brc_consensus_timeout still receives slice_id is preserved verbatim — good, since that wiring is unchanged.

Nothing blocking, no non-blocking suggestions.

— Authored by egg

[james-in-a-box]

egg review completed. View run logs

1 previous review(s) hidden.