docs: fix start_phase and complete_phase descriptions [doc-updater] - #1953
Conversation
…-cli ref Correct two factual errors introduced when #1944 updated the server-side behavior without updating the docs: - start_phase does NOT spawn agents (spawning is driven by _run_pipeline loop); docs said "spawns agents" - complete_phase does NOT advance the pipeline; add note that callers must invoke advance_phase next, and document the new current_phase field in the response Authored-by: egg
There was a problem hiding this comment.
Verified all three changes against the implementation in orchestrator/routes/phases.py:
start_phasedescription: Confirmed — the handler (lines 510–584) setsphase_execution.status = PipelineStatus.RUNNINGand returns. No agent spawning logic. The docstring at line 529 explicitly states "does NOT spawn agents."complete_phasedescription: Confirmed — the handler (lines 662–842) setsphase_execution.status = PipelineStatus.COMPLETE, does not touchpipeline.current_phase, and returns bothcurrent_phase(unchanged) andnext_phase(suggested transition) in the response (lines 814–822). Comment at line 817 references #1940.- Recovery workflow comment: Correctly updated to match.
All three changes are factually accurate.
Non-blocking suggestion: The same stale descriptions exist in docs/guides/sdlc-pipeline.md (lines 773–774):
- `start_phase` — start a phase that is in state but has no running containers
- `complete_phase` — manually mark a stuck phase as complete
Since the whole point of this PR is correcting these misconceptions, it would be worth fixing these in the same pass. The start_phase line in particular still implies "starting" means spawning work, and complete_phase says "stuck phase" which implies it unblocks transitions (it doesn't — you need advance_phase for that).
— Authored by egg
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
There was a problem hiding this comment.
Agreed. Fixed the stale start_phase and complete_phase descriptions in docs/guides/sdlc-pipeline.md (lines 773–774) to match the corrected behavior:
start_phase— mark a phase RUNNING (does not spawn agents)complete_phase— mark a phase COMPLETE (does not advance the pipeline)
Pushed in 382269c.
— Authored by egg
This comment has been minimized.
This comment has been minimized.
There was a problem hiding this comment.
Re-reviewed at 382269c. Delta since last review: one commit fixing the stale start_phase / complete_phase descriptions in docs/guides/sdlc-pipeline.md.
Verification of previous feedback: The non-blocking suggestion from the prior review (fix stale descriptions at lines 773–774 of sdlc-pipeline.md) has been properly addressed. The new wording is consistent with the corrections already made in docs/reference/orchestrator-cli.md.
New commit review:
The change at docs/guides/sdlc-pipeline.md lines 773–774 is correct:
start_phase— now says "mark a phase RUNNING (does not spawn agents)" — matches the implementation atorchestrator/routes/phases.py:529which only flipsphase_execution.statustoRUNNING.complete_phase— now says "mark a phase COMPLETE (does not advance the pipeline)" — matches the implementation atorchestrator/routes/phases.py:693which only flips status toCOMPLETEand returnscurrent_phase(unchanged) +next_phase(suggestion).
The remaining start_phase references in sdlc-pipeline.md (lines 1086+) refer to the config.start_phase pipeline parameter (which phase to begin execution at), not the MCP tool — correctly left untouched.
No other doc files (agent-tools.md, agent-recovery.md, etc.) contain stale references to these tools.
No issues found. Both commits are factually accurate documentation corrections.
— Authored by egg
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
…e conflict in orchestrator-cli.md
Conflict Resolution SummaryResolved merge conflicts with
Details on
Verification: — Authored by egg |
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
Conflict Resolution SummaryMerged The two new commits from
Verification:
— Authored by egg |
There was a problem hiding this comment.
Re-reviewed at fc42a1e. Delta since last review (382269c): one merge commit resolving conflicts with main.
Previous feedback verification: The non-blocking suggestion from the prior review cycle (fix stale start_phase/complete_phase descriptions in sdlc-pipeline.md) was addressed in 382269c and survives the merge intact at lines 773–774.
Merge conflict resolution review:
The conflict was in docs/reference/orchestrator-cli.md where main had added content from #1950 (advance_phase auto-populate note) and #1951 (phase lifecycle error reason codes) while the PR had corrected the start_phase and complete_phase descriptions. The resolution correctly:
advance_phaserow (line 230): Took main's version with the auto-populate note from #1950. Correct — the PR didn't modify this row.start_phaseandcomplete_phaserows (lines 231–232): Kept the PR's corrected descriptions. Correct.- Error reason codes section (lines 247–267): Brought in from main (#1951). Correct — no conflict with the PR's changes.
advance_phaseparameter description (line 239): Took main's version with the auto-populate detail. Correct.- Recovery workflow step 3 comment (lines 281–282): Took main's updated comment from #1950 ("automatic when advancing from plan; needed for other phase transitions where the plan was set up externally"). Correct — more accurate than the PR's original version.
- Recovery workflow step 4 comment (line 287): Kept the PR's corrected version ("Mark the phase running (does not spawn agents — the _run_pipeline loop handles that)"). Correct — main still had the stale "Start the phase (spawn agents)" text here.
The sdlc-pipeline.md auto-merged cleanly, picking up main's addition at line 854 (advance_phase auto-populate note from #1950) alongside the PR's corrections at lines 773–774.
Verification against implementation: Confirmed start_phase handler (phases.py:529) only flips status to RUNNING, complete_phase handler (phases.py:689–692) only flips status to COMPLETE without advancing current_phase, and advance_phase auto-populates the contract when leaving plan phase (phases.py:428–450). All documentation matches.
Stale reference check: No remaining stale start_phase/complete_phase descriptions found in the docs directory. The config.start_phase references in sdlc-pipeline.md (lines 1086+) are the pipeline configuration parameter, not the MCP tool — correctly left untouched.
No issues found. The merge resolution is correct and preserves all PR fixes while cleanly incorporating main's additions.
— Authored by egg
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
There was a problem hiding this comment.
Re-reviewed at 566c4c0. Delta since last review (fc42a1e): one merge commit bringing in main (primarily #1882 gateway auto-filter changes). No PR-authored changes, no conflicts — the doc corrections in orchestrator-cli.md and sdlc-pipeline.md survive intact.
No agent-mode design concerns. Documentation-only PR with no impact on agent workflows, prompts, or LLM interactions.
— Authored by egg
This comment has been minimized.
This comment has been minimized.
There was a problem hiding this comment.
Re-reviewed at 566c4c0. Delta since fc42a1e: one merge commit bringing in main (includes #1957, #1937).
Merge verification: The merge auto-merged cleanly — no conflicts in the PR-touched files (orchestrator-cli.md, sdlc-pipeline.md). The PR's corrections at lines 231–232 and 287 of orchestrator-cli.md, and lines 773–774 of sdlc-pipeline.md, survive intact.
Implementation verification: Re-confirmed all three claims against orchestrator/routes/phases.py:
start_phase(lines 510–584): only flipsphase_execution.statusto RUNNING; docstring at line 529 explicitly states "does NOT spawn agents"complete_phase(lines 662–842): only flips status to COMPLETE; does not touchpipeline.current_phase; response includescurrent_phase(unchanged) andnext_phase(suggested transition) at lines 816–822advance_phase(lines 410–459): auto-populates contract when leaving PLAN phase
Stale reference check: Remaining start_phase references in sdlc-pipeline.md (lines 1086+) are all the config.start_phase pipeline parameter, not the MCP tool — correctly untouched.
No issues found.
— Authored by egg
|
egg review completed. View run logs 11 previous review(s) hidden. |
Update
docs/reference/orchestrator-cli.mdto correct factual errorsintroduced when #1944 updated phase tool behavior without updating docs.
Changes:
start_phaseentry: removed "spawns agents" claim — this endpoint onlyflips
phase_execution.statusto RUNNING; agent spawning is handled bythe
_run_pipelineloop. The comment in the recovery workflow exampleis updated to match.
complete_phaseentry: clarified it does NOT advance the pipeline;operators must call
advance_phasenext. Added mention of the newcurrent_phasefield in the response (added in Fix #1940: make phase MCP tool descriptions name their state-machine effects #1944 to make the"no advance happened" guarantee explicit).
Triggered by: #1944
Test Plan
implemented behavior in
orchestrator/routes/phases.pyAuthored-by: egg