メニュー
Manage tasks, state, and memories - parse PRDs into hierarchical tasks with dependencies and complexity
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Use when asked to implement, build, add, fix, refactor, or change anything in a codebase — vague asks included. Tier-driven phases: intake, investigate, plan, build, verify, deliver. Entry point: /implement.
The map-reduce + blackboard pipeline that lets a small-context model digest inputs far beyond one attention window — a codebase, a long document, a corpus, a multi-part idea — and state only evidence-anchored, citation-checked conclusions. Use when / Trigger: executing or resuming an analyst run (/analyst:digest, /analyst:resume, /analyst:show), or any analysis input exceeds one 200-line chunk.
Use whenever a change needs a system-design decision — architecture, system design, data model, module/service boundaries, interface/API contracts, scale, or an ADR. The canonical discipline for designing systems: tradeoffs made explicit, the design sized to the 15 elicited context fields, one-way doors found by a five-flag checklist (one ADR each), every component earning its place through a two-cell subtraction walk, and every "start simple" seam carrying a concrete extraction trigger. Executed as checkpointed run steps (references/phase-*.md) that write into docs/architecture/ + docs/adr/ per scribe's documentation contract.
Stay focused on the project with the task in hand; recall another project's knowledge ONLY when the task actually needs it. When a task requires changing a shared dependency you own (a local composer/npm package in another repo), make the change IN that owning repo — commit, push, retag — then update and re-verify the consumer. Trigger: an edit is about to land inside a dependency directory, or a task can't be done without touching a package you maintain elsewhere.
Use when asked for a deep audit, production-readiness review, architecture review, or prioritized recommendations across a codebase or subsystem — orchestrates parallel reader agents with adversarial verification and publishes findings to docs/deep-analysis/
Run a large, multi-place task as a checkpointed state machine over the runs store — plan, adversarial plan review, wave-by-wave execution, synthesis — one step per invocation, resumable after any interruption. Trigger: cross-file refactor, data migration, large feature build, multi-source research — work that cannot finish in one pass. Manual invocation: /deep-work.
| name | taskmanager |
| description | Manage tasks, state, and memories - parse PRDs into hierarchical tasks with dependencies and complexity |
| allowed-tools | Read, Write, Edit, Glob, Grep |
.taskmanager/taskmanager.db (SQLite, schema v5.0.0) is the source of truth for all
tasks, state, and memories. You dispatch; the work lives in ONE reference file per
step. Load only the row that matches the situation — never several at once.
| Situation | Load |
|---|---|
| Parse a PRD (file / folder / prompt) into planning input | references/phase-1-parse-prd.md |
| Analyze the PRD (risks, ambiguities, macro questions) | references/phase-2-analyze.md |
| Score complexity / priority / MoSCoW / business value | references/phase-3-complexity.md |
| Define milestones + generate & insert the task tree | references/phase-4-decompose.md |
Expand tasks into subtasks (auto-loop or --expand) | references/phase-5-expand.md |
| Execute a task (select, deps, deferrals, memories) | references/phase-6-execute.md |
| Post-execution review (memories, deferrals, scribe) | references/phase-7-post-review.md |
Gate before done — acceptance criteria | references/phase-8-gate-criteria.md |
Gate before done — regression (always runs) | references/phase-9-gate-regression.md |
| Finish a task (propagation, archival, milestone) | references/phase-10-finish.md |
Verify one task standalone (verify [id], --batch, --override) | references/verify-task.md |
Verify a milestone (verify --milestone) | references/verify-milestone.md |
Verify the PRD (verify --prd) | references/verify-prd.md |
| Data contract (tables, views, valid values, IDs, logs) | references/data-contract.md |
| Multi-step checkpointing / resume | references/run-state.md |
| Verifier votes, probes, aggregation, verdict.json | references/ensemble-dispatch.md |
| A worked PRD→tasks example | references/ingest-cases/index.md — pick ONE case |
RULES (cross-phase, always):
SELECT * FROM v_next_task LIMIT 1; (or v_next_task_sequential) — use the view, never paste its logic..taskmanager/taskmanager.db except through these references' SQL.taskmanager-memory skill.runs/run_steps after EVERY step; resume = re-read state.