# Task Analysis

> Analyze an input task, issue, or feature request against the existing codebase, persist the task framing under `tmp/tasks/<task-name>/task.md`, write one or more `solution_N.md` option files, align evaluation criteria with a human when tradeoffs exist, and recommend a path with explicit reasoning.

- Skill: `blockscout/task-analysis` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add blockscout/task-analysis`
- Raw SKILL.md: https://api.skillmd.com/api/skills/blockscout/task-analysis/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Productivity
- Author: blockscout (https://skillmd.com/u/blockscout)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/blockscout/task-analysis

---


# Task Analysis Skill

Use this skill when the user wants a pre-implementation review of a task, issue, feature, or design direction.

## Workflow

Follow the canonical workflow in `../../../.memory-bank/workflows/task-analysis.md`.

## Required Guardrails

- Read the relevant `.memory-bank/` context and source-of-truth code paths before proposing solutions.
- Ground every option in the actual repo structure and current abstractions.
- Persist each analysis under `tmp/tasks/<task-name>/` using `task.md`, `solution_N.md`, and `solutions.md` when multiple options exist.
- If multiple viable options exist, explicitly align on evaluation criteria with the human before selecting a recommendation.
- If only one realistic option exists, say so and explain why.
- Separate facts, assumptions, and recommendations.

## Minimal Starting Reads

Start with:

- `../../../.memory-bank/project-context.md`
- `../../../.memory-bank/architecture.md`
- `../../../.memory-bank/exploration-map.md`
- `../../../.memory-bank/gotchas.md`

Then read the specific research notes, ADRs, rules, tests, and code paths relevant to the task under review.