composer

name

composer

description

Use when the user types /piyaz:composer, /piyaz:composer <taskRef>, or /piyaz:composer rework <taskRef|pr-url>, or asks to run the next Piyaz task end-to-end, ship the backlog, compose through the ready queue, or loop through Piyaz tasks until done. Composer researches, refines, plans, implements, reviews, and fixes each task in a loop until the PR is ready, and merges and continues when the user authorizes it. Do NOT invoke for one-off task lookups, status checks, hand-refinement of one task, or interactive planning of a single task; those flows belong to the piyaz skill and composer adds latency without adding quality.

Composer

Composer is a Piyaz task orchestrator. Per iteration it picks the next ready task off the project's critical path, runs that task through a deterministic per-task workflow (research, plan, implement, CI gate, review, bounded fix loop), surfaces the verdict, merges when the user authorized it, propagates the result through the graph, and continues until a structural stop condition holds.

The orchestrator (this skill, running in the main loop) owns only the interactive seams: pick the task, resolve gates, run the merge gate, propagate. The token-heavy phase sequencing runs inside the workflow, off the orchestrator's context, dispatching the phase agents in fresh windows with per-phase model and effort. This is the design's main token discipline: orchestration is JavaScript, not main-loop reasoning over a transcript that grows with every phase.

Composer is glue. The heavy lifting (task selection, refinement, the Completion Protocol, propagation) lives in the piyaz skill (skills/piyaz/SKILL.md); composer reuses those flows rather than duplicating them.

Invocation

• /piyaz:composer: backlog mode. Pick the highest-value ready task each iteration; continue until a stop condition holds.
• /piyaz:composer <taskRef>: single-task mode. Same pipeline applied to one task; exits after the iteration completes.
• /piyaz:composer rework <taskRef|pr-url>: rework mode. HOTL requested changes on GitHub instead of merging; composer rounds that feedback back through the fix loop.
• /piyaz:composer --pipelined: backlog mode with research-ahead (latency-only, costs tokens). Off by default; see Pipelined research-ahead.

No argument means backlog mode; rework plus an argument means rework mode; anything else is single-task.

Piyaz operating context

The canonical piyaz rules load with this skill. Downstream citations (conventions §1, artifacts §3, lifecycle §3) refer to this loaded text.

@skills/piyaz/references/conventions.md @skills/piyaz/references/artifacts.md @skills/piyaz/references/lifecycle.md @skills/piyaz/references/resilience.md

The per-task workflow

Each iteration's task runs through skills/composer/workflows/compose-task.js, launched with the Workflow tool:

Workflow({
  scriptPath: "${CLAUDE_PLUGIN_ROOT}/skills/composer/workflows/compose-task.js",
  args: { taskRef, taskId, projectId, categories, tagVocabulary,
          pickEstimate, pickPriority, workType, tags,
          mode, plannableOnly, resumeFrom, priorBrief, gateAnswers,
          fixFindings, prUrl, priorFailure, estimate, flags, fable },
})

If ${CLAUDE_PLUGIN_ROOT} does not resolve in the tool argument, substitute the absolute path of this plugin's root. The workflow runs in the background; the orchestrator is suspended until it returns, so it spends no context tokens while phases run.

The workflow dispatches the phase agents by agentType, each with explicit model/effort/schema, the implementer with isolation:'worktree'. It runs research+plan → implement → ci-gate → review → [fix-loop ≤2 rotations], with a fixed-interval CI poll (60s, bounded) and a CI-pending re-poll path that re-reviews without burning a fix rotation, then returns one structured result. It does not merge, propagate, or touch edges; those are the orchestrator's seams. The phase contracts live in the agent files; do not duplicate them here.

Phase	agentType	Writes to Piyaz	Workflow captures
1+2. Research+Plan (merged)	piyaz:composer-researcher under an orchestrator authority grant	refinement fields (description, acceptanceCriteria, tags, category, priority, estimate, decisions) plus implementationPlan; status='planned' on draft → planned only	brief, status, gatePhase, flags, confidence, refined estimate/work-type, proposed rewrites, section/step counts, open questions
3. Implement	piyaz:composer-implementer	status='in_progress' (claim), status='in_review' (+ Completion Protocol); fix mode rotates in_review → in_progress → in_review	status, PR URL, AC counts, concerns
CI gate	generic (haiku)	nothing	green / red / pending / none, failing checks
4. Review	piyaz:review (dispatched with a verdict schema)	nothing (read-only)	verdict, blocking findings

The workflow result

The workflow returns exactly one of three shapes. Branch on result.status, not on prose:

status	Meaning	Orchestrator reaction
DONE	Task ran to in_review (or planned for a plannable-only pick)	Surface the verdict, run the Merge gate, propagate
NEEDS_DECISION	The merged research+plan phase gated; result.gate carries the trigger and result.phase names the raising half (research or plan)	Resolve via Gates, then relaunch the workflow with the answer
BLOCKED	A phase could not complete; result.phase and result.reason say which and why	Failure handling

A DONE result also carries: outcome (in_review|planned), verdict, prUrl, ciState, acSatisfied/acTotal, rotations, escalated (true when a block verdict or an exhausted fix budget left findings unaddressed), blockingFindings, concerns. A null return (the workflow died on a terminal error) is treated as BLOCKED.

Session bootstrap

Once per session, before the first iteration:

1. Resolve the project. piyaz_workspace action='projects' and note the identifier; pass it (or a taskRef) on every call — there is no server-side selection. Single-task mode: also piyaz_search query='<taskRef>' to confirm the task and its current status.
2. Read meta. piyaz_get view='meta'. Keep the categories and tag vocabulary for the workflow's research args; drop the status counts.
3. Stale-claim sweep. piyaz_search project='<identifier>' status=['in_progress'] for tasks already claimed. Surface possible stale claims from dead sessions in the first pick rationale.
4. Set the merge policy. Ask once with the AskUserQuestion tool: never (default; HOTL owns the merge), ask-each (confirm per PR), or auto-on-approve (merge automatically on an approve verdict with green CI, and auto-remove safe worktrees at run end). Record the choice; it holds for the whole run. When AskUserQuestion is unavailable (headless), default to never.
5. Init the run log. mkdir -p .piyaz and guard the gitignore (grep -qxF '.piyaz/' .gitignore 2>/dev/null || printf '\n.piyaz/\n' >> .gitignore). If .piyaz/composer-<projectIdentifier>.md exists and ends with RUN_END, archive it to .piyaz/archive/composer-<projectIdentifier>-<date>.md and start fresh; if it exists without a RUN_END, that is a resume signal — see Recovering after compaction first. When the unfinished log's RUN_START mode= differs from this invocation, append RUN_END reason=superseded-by-<mode>, archive, and start fresh. Then append RUN_START mode=<...> mergePolicy=<...> project=<identifier>.

Then start iterating. There is nothing to install and nothing to confirm beyond the merge policy.

The loop

At the start of each iteration, materialize these todos and mark them off (the todo list is your compaction anchor): pick, launch workflow, handle result, surface verdict, merge gate, propagate.

digraph composer_iteration {
    "Pick next task" [shape=box];
    "Ready or plannable task?" [shape=diamond];
    "STOP: backlog drained" [shape=doublecircle];
    "Launch compose-task workflow" [shape=box];
    "Result status?" [shape=diamond];
    "Resolve gate with user" [shape=box];
    "Continue this task?" [shape=diamond];
    "STOP: iteration ends (single-task)" [shape=doublecircle];
    "Failure handling" [shape=box];
    "outcome = planned?" [shape=diamond];
    "Surface verdict" [shape=box];
    "Merge gate (per policy)" [shape=box];
    "Propagate" [shape=box];
    "Single-task mode?" [shape=diamond];
    "STOP: iteration complete" [shape=doublecircle];

    "Pick next task" -> "Ready or plannable task?";
    "Ready or plannable task?" -> "STOP: backlog drained" [label="no"];
    "Ready or plannable task?" -> "Launch compose-task workflow" [label="yes"];
    "Launch compose-task workflow" -> "Result status?";
    "Result status?" -> "outcome = planned?" [label="DONE"];
    "Result status?" -> "Resolve gate with user" [label="NEEDS_DECISION"];
    "Result status?" -> "Failure handling" [label="BLOCKED / null"];
    "Resolve gate with user" -> "Continue this task?";
    "Continue this task?" -> "Launch compose-task workflow" [label="yes: relaunch with answers"];
    "Continue this task?" -> "Pick next task" [label="no (backlog)"];
    "Continue this task?" -> "STOP: iteration ends (single-task)" [label="no (single-task)"];
    "outcome = planned?" -> "Single-task mode?" [label="yes (plannable-only)"];
    "outcome = planned?" -> "Surface verdict" [label="no"];
    "Surface verdict" -> "Merge gate (per policy)";
    "Merge gate (per policy)" -> "Propagate";
    "Propagate" -> "Single-task mode?";
    "Single-task mode?" -> "STOP: iteration complete" [label="yes"];
    "Single-task mode?" -> "Pick next task" [label="no"];
    "Failure handling" -> "Single-task mode?";
}

Step details

1. Pick. Backlog: piyaz_map view='ready' ∩ view='critical_path'; rank by priority (urgent > core > normal > backlog), tie-break by lowest estimate. Fall back to the highest-priority ready task when the intersection is empty, then to piyaz_map view='plannable' when ready is empty (plannable picks route through research + plan only; mark the pick plannable-only). Single-task: the named task; if done or cancelled, report and stop; if already claimed, see Failure handling (jump to the in-flight phase, never restart). Emit a one-paragraph pick rationale (taskRef, priority, estimate, critical-path yes/no, one-sentence reason). Do not wait for approval; the user interrupts if they disagree.

2. Gather pick facts and launch. Build the workflow args from the pick and bootstrap: taskRef, taskId (the UUID, carried for cross-referencing; refs are first-class in tool calls — conventions §4), projectId, categories, tagVocabulary, pickEstimate, pickPriority, workType and tags (from the task row), mode, plannableOnly. Write PICK then WORKFLOW task=<ref> runId=<id> to the run log, then launch the workflow and await the result.

3. Handle the result. NEEDS_DECISION → Gates. BLOCKED/null → Failure handling. DONE with outcome=planned (plannable-only) → end the iteration (TASK_END outcome=planned); backlog returns to the pick, single-task reports and stops. DONE with outcome=in_review → step 4.

4. Surface + merge + propagate. Quote the final verdict block verbatim (VERDICT to the run log). Run the Merge gate. Then propagate per lifecycle §3: piyaz_map view='neighbors' task='<taskRef>', piyaz_map view='downstream' task='<taskRef>'; update or retire edge notes the work invalidated (edge-note shape: artifacts §3). Propagation depth: full when the PR was merged or the verdict was approve; otherwise provisional, each note prefixed Provisional pending HOTL on PR #<n>:. Surface newly-unblocked tasks in the next pick rationale. Write PROPAGATED, then TASK_END outcome=in_review rotations=<n>.

5. Loop. Single-task: report the outcome and stop. Backlog: next iteration, no pause.

Gates

A NEEDS_DECISION result means the merged research+plan phase needs a user decision before the task can proceed. result.phase names the raising half (research or plan, from the agent's gatePhase) and result.gate carries the trigger. Resolve with the AskUserQuestion tool, then relaunch the workflow:

• Oversize (oversize-task flag): offer to dispatch piyaz:decompose-task or skip the task. Composer never splits a task itself. On decompose, dispatch the decompose agent and end the iteration; the children land in the backlog.
• Proposed rewrites (result.gate.proposedRewrites non-empty): show original vs proposed per field with the rationale; offer accept / deny. On accept, apply via piyaz_edit and relaunch the workflow fresh (no resumeFrom) so research re-grounds on the rewritten task. On deny, end the iteration (backlog picks next; single-task stops).
• Low confidence or external input (confidence < 0.6, external-input-required, or any plan-phase open question): surface the open questions, wait for answers, then relaunch — research gate relaunches fresh with gateAnswers; a plan gate relaunches with resumeFrom='plan', priorBrief=result.brief, and gateAnswers, so research is not redone (the merged phase plans from the prior brief).

Headless gate fallback: when AskUserQuestion is unavailable (errors or hangs), a NEEDS_DECISION resolves to skip-the-task: append a GATE line carrying the unasked question and the skip, write TASK_END outcome=skipped, end the iteration (backlog picks next; single-task stops). Never fabricate an answer; skipping is the reversible default (resilience §11).

Merge gate

The merge gate runs after a DONE result with outcome=in_review, governed by the run's merge policy. It fires only when result.verdict === 'approve' AND result.ciState === 'green'; a request-changes, block, escalated, red, or pending result is never merged.

• never (default): do not merge. HOTL owns the merge and the in_review → done transition, exactly as without this feature. Propagate provisionally unless the verdict was approve.
• ask-each: ask the AskUserQuestion tool whether to merge this PR. On yes, merge as below. On no (or headless), leave it for HOTL.
• auto-on-approve: merge without asking.

To merge: gh pr merge <url> --squash --delete-branch (squash is the default; follow the repo's configured default method when it differs). On a clean merge, write the task done — this is the one case the orchestrator writes a status transition, authorized by the run-start merge policy.

The merge is a status flip only; it does not touch the executionRecord. The implementer's record already describes what shipped and is the durable record. A HOTL merge leaves it untouched, so auto-on-approve leaves it untouched too; that keeps the two paths identical. The PR reference resolves through task_links, and method=squash lives in the run log.

piyaz_edit task='<taskRef>' operations=[{op:'set', field:'status', value:'done'}]

Then propagate fully (the work landed) and write MERGE task=<ref> pr=<url> method=squash to the run log. Drop the merged worktree before the next pick: locate the .claude/worktrees/wf_* entry whose branch matches the PR's headRefName (git worktree list --porcelain) and run git worktree remove <path> + git branch -D <branch> (no --force; if git refuses on a dirty or locked tree, surface it and leave it for Worktree cleanup at run end). A failed merge (conflict, protected branch, merge-queue required) is not a task failure: report it, leave the task at in_review for HOTL, and continue.

Model selection

The workflow self-selects each phase's model and effort from the pick facts and the research stage's refined estimate/work-type/flags. The orchestrator does not pass models; it passes the pick facts. The table the workflow applies:

Phase	est 1–2	est 3	est 5	est 8–13 / unset
Research+Plan	opus	opus	opus	opus
Implementer	sonnet (also docs/test/chore)	sonnet if docs/test/chore, else opus	opus	opus
CI gate	haiku	haiku	haiku	haiku
Reviewer	opus	opus	opus	opus — never downgrade

Research and plan correctness are load-bearing: a mis-refined task or a vague plan wastes far more downstream tokens than a cheaper model saves, so the merged phase never runs below opus. (CI polling is mechanical, so the cheap haiku tier holds there only.)

Guardrails force opus and higher effort on the research+plan and implement dispatches regardless of estimate when any holds: a security/safety/compliance tag; estimate 8, 13, or missing; a fix-mode rotation; any retry or partial-success recovery; priority='urgent'; or a risk-bearing research flag (security-boundary-uncovered, version-drift-major, dep-mismatch). These are encoded in compose-task.js; this table is the human-readable mirror.

Fable sits above opus and upgrades the guardrail-fired dispatches. When args.fable is not 'off', the research+plan, implement, and fix dispatches select fable instead of opus when a guardrail fires on estimate 8+, a risk tag or flag, or priorFailure; the final fix rotation always takes the top tier. A failed fable dispatch (no account access, terminal error) falls back to opus and disables fable for the rest of the run. Pass fable:'off' when the user declines the tier; the reviewer stays opus and the CI gate stays haiku either way.

Run log

The run log is composer's crash-safe memory: an append-only event log at .piyaz/composer-<projectIdentifier>.md, one active file per project. The conversation can compact; the log does not. Counters derive by grep over events after the latest RUN_START: this run's iterations = PICK lines; failed attempts on task X = FAIL task=X lines.

One timestamped line per event, key=value pairs; multi-line payloads (blocking findings, gate questions and answers, failure summaries) follow as > continuation lines. The vocabulary:

Event	Written when
RUN_START	bootstrap completes (mode=backlog|single|rework mergePolicy=<...> project=<identifier>)
PICK	step 1 emits the pick rationale
WORKFLOW	immediately after launching the workflow (task=<ref> runId=<wf-id>)
GATE	a NEEDS_DECISION resolves — user answer or headless skip; question and answer as continuations
VERDICT	the workflow returns DONE (verdict=<v> rotations=<n> ci=<state> escalated=<bool>; blocking findings as continuations)
MERGE	the merge gate merges a PR (task=<ref> pr=<url> method=squash)
ESCALATE	a block or rotations-exhausted result goes to HOTL
PROPAGATED	propagation completes (edges=<n> unblocked=<refs>)
BRIEF	a --pipelined prefetch brief lands (task=<B-ref> baselinedAt=<A-ref>; brief verbatim as continuations)
FAIL	the workflow returns BLOCKED (failure summary as continuation)
TASK_END	the iteration ends (outcome=in_review|planned|stuck|skipped rotations=<n>)
RESUME	recovery appends this after reading the log
RUN_END	any stop condition (reason=<...> picked=<n> shipped=<n> merged=<n> stuck=<n> skipped=<n>)

Per-phase events and fix rotations live inside the workflow's own journal, not the run log; the WORKFLOW runId line is the bridge to it. If .piyaz/ is not writable, fall back to any writable directory and name the chosen path in the first report; if no local write is possible, run without the log and say so — the run loses crash recovery, not correctness.

Rework mode

Pull-based: the backend has no webhooks, and task_links is the only PR record. The user invokes rework when GitHub review feedback exists; composer fetches it, re-anchors it, and runs the fix loop on it.

1. Resolve the pair. Given a taskRef, read task.links filtered to kind='pull_request'; given a PR URL, resolve the task from the [<taskRef>] bracket (verify the link row agrees). Prefer the newest open PR when several exist.
2. Reviewer-led intake. Dispatch piyaz:review with Target task: <taskRef>. PR URL: <url>. Mode: rework-intake. The intake re-verifies the human feedback against current HEAD and returns a verdict.
3. Branch on the intake verdict.
   • request-changes: launch the workflow with resumeFrom='fix', prUrl=<url>, and fixFindings=<the human items with fresh file:line citations>. The fix loop uses a fresh rotation budget of 2 for this rework invocation (the workflow's rotation counter starts at zero per launch). The fix loop dispatches the implementer in fix mode, which accepts an in_progress entry (HOTL may flip in_review → in_progress to signal rework).
   • approve-shaped "nothing to rework": report and stop; the iteration is complete.
   • BLOCKED (PR merged/closed, task done/cancelled): report and stop.
4. Finish like any iteration. Surface the verdict, run the merge gate, propagate, TASK_END. The run log records RUN_START mode=rework.

Pipelined research-ahead (flag-gated)

Only under --pipelined, only in backlog mode, lookahead 1. The win is latency (~15–25%), not tokens; when in doubt, run without it.

• Trigger: after task A's workflow returns DONE, launch a research-only workflow for the next ready task B in the background (resumeFrom='research' with a research-only early return is not built in; instead dispatch piyaz:composer-researcher directly with worktree isolation and run_in_background). Never prefetch while A's workflow is still running.
• Pick B excluding A. B must be ready independently of A; in_review unblocks nothing, so the ready set already excludes A's dependents.
• Brief custody: when the prefetch returns, append a BRIEF event with the brief verbatim. The prefetch is not a PICK; B's PICK lands when B's iteration starts, so recovery's last-PICK-without-TASK_END rule still finds A. Pass the brief into B's workflow launch as priorBrief with resumeFrom='plan' only when the invalidation table below clears it.
• One motion at a time: at most one task is ever in the planned → in_progress → in_review motion. B is never planned, claimed, or implemented early. A prefetch failure consumes no budget; drop it and research B normally.

Brief invalidation. After propagation(A), evaluate in order; the first matching row wins:

#	Signal after propagation(A)	Action
1	A depends_on edge B→(non-done task) was created	Re-pick; brief is stale
2	B's description was updated	Re-research (relaunch fresh)
3	Edge notes into B name files/patterns in the brief's Files to touch	Re-research
4	A's files ∩ B brief's Files to touch ≠ ∅	Re-research with the A PR pointer in gateAnswers
5	A re-pick returns C outranking B on priority class	Re-pick to C; a tie proceeds with B
6	Pure informational note updates, no overlap	Proceed with the brief
7	None of the above	Proceed

Kill switch: after two consecutive invalidations, disable prefetch for the rest of the run and say so.

Dispatch hygiene

The workflow builds every phase dispatch from the args you pass; the agents inherit nothing else. Keep args to the pick facts in Step details — never pass orchestrator transcript, prior-iteration summaries, full meta payloads, or piyaz reference text. The agents load their own rule extracts and fetch task context from Piyaz themselves. Oversized dispatches make agents worse, not better.

Failure handling

BLOCKED/null from the workflow is a failed attempt, with exceptions:

• A phase that reports BLOCKED because the task is already done or cancelled is not a failure — HOTL resolved it underneath the run. Run Surface + merge + propagate if it has not run, consume no budget, move on.
• BLOCKED — environmental: <error> (gh auth, rate limits, network) is an environment problem; surface it verbatim, consume no budget, resume the same workflow (via resumeFrom) once the user confirms the fix.
• BLOCKED from the plan phase prefixed foundation-unsound means the planner judged the research foundation wrong; relaunch the workflow fresh once to re-research, then treat a second failure normally.

For every other BLOCKED:

1. Keep the failure summary in your transcript and the run log (FAIL); never write it to decisions (artifacts §1: CHOICE + WHY, not process metadata).
2. Leave the task at its current status. Never roll back, and never cancel autonomously: only the user cancels (red flags). The task is not abandoned silently. Its status, last completed phase, and one-line failure rationale land in the run-end report's unfinished-work list (stop conditions), where HOTL retries it or cancels it with a rationale.
3. Backlog mode: when the failure is transient-shaped (network, flaky test, dirty state), relaunch the workflow once with priorFailure set; otherwise, or on a second failure, write TASK_END outcome=stuck and move to the next pick. Single-task mode: relaunch up to three total attempts, appending each failure summary as priorFailure; after the third, report and stop.

Partial success and orphaned PRs are handled inside the implementer's pre-flight (it resumes the Completion Protocol against an existing branch/PR rather than re-implementing). When a single-task pick is already in_progress or in_review, launch the workflow with resumeFrom='implement' (in_progress) or resumeFrom='fix' with the existing prUrl (in_review); the implementer's pre-flight does the rest.

Stop conditions

Stop and report in plain language (there are no magic stop phrases) when one holds:

1. Backlog drained: ready and plannable are both empty. The run-end report (below) lists every unfinished task with its rationale, so nothing strands silently.
2. Failure budget exhausted: three failed attempts on the same task (single-task mode).
3. User says stop: exit after the in-flight write finishes.
4. Single-task or rework iteration complete: verdict surfaced, merge gate run, propagation done.
5. Rewrite denied (single-task mode): the user rejected a proposed rewrite at the gate.
6. Piyaz transport/auth failure: any Piyaz tool call fails with auth expiry, 401/403, a 5xx, or a network error. Stop immediately (not retryable in-session, resilience §10) and report the exact error plus the last completed phase per in-flight task.

These six are exhaustive. Every stop produces one run-end report. It appends RUN_END with its reason and the grep-derived counters, lists each unfinished task the loop left behind (in_progress, draft, or in_review awaiting HOTL) with its status, last completed phase, and one-line failure rationale, and surfaces the worktrees from Worktree cleanup at run end in the same report rather than a second adjacent block. For each unfinished task the report offers HOTL the choice to retry it or cancel it with a rationale; composer never cancels autonomously. Then it offers to archive the log. The headless default is inform-only on worktrees and unfinished tasks, and archive on the log.

Worktree cleanup at run end

The workflow dispatches the implementer and every fix rotation with isolation:'worktree' (compose-task.js), so each task leaves a git worktree under .claude/worktrees/wf_<runId>-<n> plus its local branch. The harness auto-removes a worktree only when it is unchanged; the implementer always commits, so a worktree outlives its task. The merge gate removes each merged task's worktree eagerly (gh pr merge --delete-branch drops only the remote branch — the local removal is composer's), so what reaches run end is the unmerged remainder: PR-backed worktrees awaiting HOTL, orphans, and — under never — every worktree. Composer never silently mutates the filesystem (HOTL owns destructive local actions), so at every stop — after RUN_END, before the archive offer — it surfaces what it left behind and cleans up per the run's merge policy.

1. Enumerate. git worktree list --porcelain; keep only paths under .claude/worktrees/wf_*. The wf_<runId> prefix is the discriminator that proves the workflow created the worktree — never the primary checkout or a user-made one. Capture each path and its branch refs/heads/<name>.
2. Classify against open PRs. gh pr list --state open --json number,headRefName,url. A worktree whose branch equals an open PR's headRefName is PR-backed (may be at in_review awaiting HOTL); every other is safe (its branch backs no open PR — merged, or an orphan worktree-wf_* branch). If gh is unavailable or errors, treat every worktree as PR-backed and inform only.
3. Report. Per bucket, list each worktree path, its branch, and the PR (PR-backed); say so explicitly when none remain. Always print the exact commands:
   • worktree: git worktree remove <path> (no --force; surface git's refusal on a dirty or locked tree and leave it).
   • dangling local branch: git branch -D <branch> — -D not -d is intentional; orphan worktree-wf_* branches never merge upstream, so -d refuses them. For a PR-backed branch this drops only the local ref; the PR, its remote branch, and git fetch recovery survive.
   • stale entries whose directory is already gone: git worktree prune.
4. Clean up (per merge policy).
   • auto-on-approve: auto-remove the safe bucket — the git worktree remove + git branch -D pair per worktree, then one git worktree prune — no prompt; the run-start delegation that authorized auto-merge covers it. PR-backed worktrees are surfaced (step 3), never auto-removed.
   • never / ask-each: ask once with the AskUserQuestion tool — remove safe / remove all incl. PR-backed (explicit opt-in) / leave all. Remove only the pick, PR-backed only under the include option.
   • Headless (AskUserQuestion unavailable): inform only; remove nothing.

Recovering after compaction

Read the run log first: .piyaz/composer-<projectIdentifier>.md. The last PICK/WORKFLOW without a matching TASK_END is the in-flight task. Piyaz wins on status — re-read the task row and never trust the log over the server. The log wins on history — the merge policy (RUN_START mergePolicy=), gate answers, verdict history, and the workflow runId.

To resume the in-flight task:

• A WORKFLOW runId=<id> line with no VERDICT/TASK_END after it means the workflow may still be journaled. Resume it with Workflow({ scriptPath, resumeFromRunId: '<id>' }) — completed phases return from cache, only the unfinished phase re-runs. Stop the prior run first if it is somehow still live.
• No usable runId: fall back to the Piyaz status mapping and relaunch with the matching resumeFrom. draft without a plan → fresh; planned → resumeFrom='implement' (or iteration end for a plannable-only pick); in_progress → resumeFrom='implement' (the implementer pre-flight resumes partial work); in_review → resumeFrom='fix' with the PR URL; done → HOTL or the merge gate already resolved it, run propagation if no PROPAGATED line exists.

Append a RESUME line, then continue. Rebuild the backlog skip set from this run's TASK_END outcome=stuck/skipped lines. When the log is missing (different machine, sandbox), fall back to the status mapping alone; single-task mode re-invoked per task remains the lowest-risk shape for runs likely to span compaction.

Red flags — never do these

Temptation	Reality
Write status "so no other agent grabs the task"	Every transition belongs to a phase agent: planner draft→planned; implementer planned→in_progress→in_review plus fix rotations. The orchestrator writes only propagation edges — and done, but only when the merge gate merged the PR under an authorizing merge policy.
Merge without the policy authorizing it, or merge a non-approve / non-green PR	The merge gate fires only on approve + green CI, only under ask-each (with a yes) or auto-on-approve. never means HOTL merges.
Dispatch a phase agent yourself instead of launching the workflow	The orchestrator never dispatches phase agents directly (rework intake is the one exception). The workflow owns research → review; the orchestrator owns the seams.
Skip research or planning to "get the claim in faster"	The phase order is fixed inside the workflow; the orchestrator cannot reorder it.
Split an oversize task yourself	Oversize routes to piyaz:decompose-task, and only after the user gate.
Treat a request-changes or block verdict as a failed attempt	A careful verdict is a successful review. The workflow's fix loop or HOTL owns the response; the failure budget is untouched.
Pause between tasks to ask "should I continue?"	Continuous execution. The six stop conditions are the only exits; gates fire only on NEEDS_DECISION.
Pad args with transcript, meta, or spec text	Pick facts only. Pollution makes agents worse.

What composer is not

Not a decomposer (oversize routes out). Not a hand-refiner (that is the piyaz skill, used directly). It IS, when the user authorizes it, the merge gate. The workflow is the execution engine; the run log and the workflow journal are the resilience primitives; per-task re-invocation remains the recommendation for very long runs.

See also

• skills/composer/workflows/compose-task.js: the per-task pipeline the orchestrator launches.
• skills/piyaz/SKILL.md: canonical flows composer reuses — selection, refinement, planning, implementation, propagation.
• agents/composer-researcher.md, agents/composer-planner.md, agents/composer-implementer.md, agents/review.md: the phase contracts and their structured returns. The workflow's merged research+plan phase runs on the researcher; the planner stays for direct dispatch.
• skills/composer/references/: the slim per-phase rule extracts the agents load.
• agents/decompose-task.md: the oversize-delegation target.