Solve Task

Solve Task

Gather the right context for a task, distill it into a brief, then enter the normal development workflow. This skill does NOT plan or implement — it disciplines context-gathering and hands the brief to superpowers:brainstorming (which leads to writing-plans → subagent-driven-development, or the execution strategy chosen in the startup survey (inline / subagent / lite)).

Inputs

$ARGUMENTS is either:

• a task key (e.g. PRI-4, matching the board's key_pattern), or
• a free-text description (e.g. "add a logout endpoint").

Pipeline

0. Startup: survey + Preflight (index freshness + task-corpus warm-up). Run this BEFORE anything else. First resolve, once, the repo path (git rev-parse --show-toplevel) and the working branch (git branch --show-current; if it is in REVIEW_BRANCHES use it, else the primary branch) — step 3 reuses the same branch for search_codebase.

   Resolve task_board exactly once before any board call. Read the repo .review.yml block: its task_board mapping has priority; an absent block falls back to get_board_config(); an explicit empty task_board: disables board work for this run. Keep the resolved {type, project, key_pattern, create_target, done_target, options} value and reuse this resolved value in Step 1 and every board operation. Never call the deploy fallback when the repo explicitly disables the board.

   Call prepare_task_context(repo, key, branch, warm_board=True) once here — pass the resolved key when $ARGUMENTS matches key_pattern, else board-less: pass the user's formulation verbatim as key (the tool's key parameter is required and accepts free text when no board key matches). It folds the manual round trips below into one deterministic call and returns a single payload: preflight (branch, indexed_sha, drift, summaries, chunks, graph_nodes — feeds Step 0.1/0.4 below), task_board/task (feeds Step 2), related/subsystems/code/test_exemplars (feeds Step 3), gaps (a list of {section, reason, cause, cause_detail, remedy} — branch on whether cause is unknown, not on one non-unknown value; the classes are storage_unavailable | embedder_unavailable, and cause_detail refines storage_unavailable only (auth_failed | missing_database | pool_exhausted | null); copy every entry into Constraints / open questions in the Step 4 brief verbatim) and warnings (surface to the user). No source failure raises — a failed section just adds a gaps entry, so the steps below stay fail-open.

<!-- include: solve-task/references/preflight.md -->

1. Config. Reuse the resolved value from preflight; do not read .review.yml or call get_board_config() again. If no board resolved, continue board-less. For incomplete metadata call get_board_targets(board_type=<task_board.type>, project=<task_board.project>, provider_options=<task_board.options or {}>): select from targets by label, and use option required_for / choices to ask for missing options. Never guess a target or an option and never branch on a board type.

Brief-building unit (Steps 2–4) runs on the chosen model. Steps 2–4 (identify → gather → distill → persist) are non-interactive; run them on the model chosen in the Step 0 startup survey. Dispatch a subagent on the chosen model when a per-subagent model override is available (session-less tools + Read/Bash/Glob/Write to persist the brief), passing the Step 0 prepare_task_context payload into its dispatch prompt so it does not re-call prepare_task_context or the tools it already replaced; otherwise build the brief inline. Details, the existing-artifacts pre-dispatch warn, the Собран на: marker line and the fail-open fallback are in references/brief-format.md below.

2. Identify the task.

   • If $ARGUMENTS matches the board's key_pattern:
     1. Store-first. Use payload.task from the Step 0 prepare_task_context call — it already carries the task's own normalized content ({key, aliases[], title, description, criteria[], status, url}) from the reviewer store, refreshed by that call's board warm-up. Call reviewer get_task(key, project=<task_board.project>) directly only as a fallback: when payload.task is empty without a matching gaps entry explaining why, or when prepare_task_context itself is unavailable — it returns the same normalized shape.
        • Hit (a task object with a key, from payload.task or the fallback get_task call): use it directly as the TaskBrief. The task is already indexed (the preflight sync persisted it) — do NOT call index_task. Note in the brief that the task data came from the reviewer store (after sync).
          • Thin criteria (optional, fail-open). The store can return criteria=[]; requirements normally live in description. If it has NO heading matching (?i)(критери|приёмк|acceptance), leave criteria empty and record the gap. Do NOT call index_task.
        • Miss (null / no key) AND a board is resolved: call generic incremental sync_board(board=<task_board.project or null>, board_type=<task_board.type>, provider_options=<task_board.options or {}>, limit=null, purge_orphaned=false), then retry get_task(key, project=<task_board.project>) once. Error or second miss → board-less: treat $ARGUMENTS as the task description and record the gap.
        • Miss AND no board: board-less — treat $ARGUMENTS as the task description.
   • Otherwise: treat $ARGUMENTS as the task description; do not perform external task reads.

   Store-first cuts the double-fetch: prepare_task_context's board warm-up already pulled the whole board into the reviewer store, and a miss gets one generic incremental sync/retry (fewer LLM tokens and no provider-specific client dependency).

3. Gather context (best-effort, fail-open). Any tool returning a "(… unavailable)" / "(ничего не найдено)" note or an error is non-fatal — continue. Details (subsystem prior, project scope, related-task lookups, search_codebase, graph deepening, test exemplars, lazy expansion/PR diff) are in references/context-gathering.md below.