com um clique
tasks
Generate an implementation task list from validated spec artifacts
Instalar com Codex ou Claude Copie este prompt, cole no Codex, Claude ou outro assistente e deixe que ele revise a página da skill e instale para você.
Menu
Generate an implementation task list from validated spec artifacts
Instalar com Codex ou Claude Copie este prompt, cole no Codex, Claude ou outro assistente e deixe que ele revise a página da skill e instale para você.
Baseado na classificação ocupacional SOC
A method for organizing a prototype that illustrates an agentic, spec-driven PR review—the type a person employs to determine if a merge is acceptable without scrutinizing each alteration. Applicable when a user intends to create, outline, or develop a review interface, a review dashboard, a PR-review process, or a "verdict screen" for an agent-generated modification set, particularly one generated by a specification-driven development (SDD) workflow. Triggers encompass: "review flow", "PR review prototype", "review UI", "how should I present this review", "verdict-first review", "visualize the PR", or asking how a developer should review an agent's output. This is a METHOD skill — it tells you how to think about and lay out the prototype; it does not lock you to a single output format. Do NOT use for performing the code review itself (finding bugs, judging the author) — that is a different skill.
Reviews a pull request in two stages — first interrogating the author to verify they understand and can defend every decision in their own submission (catching AI-generated code forwarded without comprehension), then performing a structural design review of intent fidelity, abstractions, coupling, and failure behavior. The comprehension gate must pass before the design review runs. Use when reviewing a feature PR, vetting incoming code for author ownership, screening for AI slop, or running a deep structural review beyond what CI and linters cover.
Interview the user one question at a time, walking the decision tree, to clarify a feature proposal before implementation
Translate the spec into a complete, traceable set of acceptance criteria in EARS form
Capture feature-level design decisions and the technical constraints that follow, traceable to acceptance criteria
Stress-test the spec pipeline outputs against each other and the codebase before implementation begins
| name | tasks |
| description | Generate an implementation task list from validated spec artifacts |
Translate a validated spec into an ordered, atomic, AC-traceable execution list an implementing agent can run task by task.
Pipeline position: proposal → spec → rules → review → tasks -> execute
You translate a validated spec into a task list written to disk. You do not write code, run tests, or modify project files outside spec/tasks.yaml. You do not ask questions; document judgment calls in decisions for the user to review.
Read spec/review.md (if exists) first. Locate the verdict line under ## Summary.
spec/tasks.yaml.source: review/MAJOR-N or a risk annotation on an existing task. Note in assumptions.Risk Hotspots from the review surface as risk annotations on the relevant task, regardless of verdict.
CLAUDE.md / AGENTS.md / GEMINI.md, build files, source treeSpec takes precedence over the proposal.
Read agent guidance files. Note package layout, module boundaries, naming, build/test/deployment patterns, and architectural style (layered, hexagonal, feature-sliced). Tasks place artifacts in paths consistent with the existing structure. Phases respect the existing architectural style unless rules.md mandates a deviation.
Pick an organizing principle and state it in organizing_principle:
State the choice and one-line reason in decisions.
coversPlace checkpoints where human review meaningfully reduces risk:
Every phase ends with a checkpoint. Intermediate checkpoints allowed within a phase.
phase-1, phase-2, ... in execution ordertask-N.M (phase number, task number)cp-N (terminal) or cp-N.M (intermediate)dec-1, dec-2, ...Aim for ≤ 5 phases, ≤ 7 tasks per phase. If you exceed:
decisions rather than padding.decisions.Don't pad or merge to fit the numbers.
Every AC in criteria.md appears in some task's covers.acs, OR in coverage_deferrals with a reason. No third option.
tasks:
feature: "<name>"
review_verdict: "<PASS | PASS WITH CONDITIONS>"
organizing_principle: "walking_skeleton | layered | feature_slice | risk_first"
assumptions:
- "<assumption to verify>"
decisions:
- id: dec-1
decision: "<judgment call>"
reason: "<why>"
alternatives: ["<alt 1>", "<alt 2>"]
coverage_deferrals:
- ac: AC-12
reason: "<why not in a task>"
phases:
- id: phase-1
name: "<phase name>"
description: "<what this accomplishes>"
covers: [AC-1, AC-2]
entry_criteria: "<what must be true to start>"
tasks:
- id: task-1.1
name: "<task name>"
description: "<what to do>"
artifact: "<file path or outcome>"
covers:
acs: [AC-1]
rules: [RULE-3, RULE-7]
depends_on: []
complexity: "S | M | L"
validation: "<how to verify>"
risk: "<from Risk Hotspots, if applicable>"
source: "<review/MAJOR-N if addressing a review finding>"
checkpoint:
id: cp-1
description: "<what to review>"
criteria:
- "<criterion 1>"
Required: top-level feature, review_verdict, organizing_principle, phases; phase id, name, description, covers, tasks, checkpoint; task id, name, description, artifact, covers, depends_on, validation; checkpoint id, description, criteria.
Optional: assumptions, decisions, coverage_deferrals, entry_criteria, complexity, risk, source.
Complete only when ALL hold:
criteria.md in some task's covers.acs or in coverage_deferralsdepends_on references an earlier task in execution orderorganizing_principle set and justified in decisionsrisk annotationsdecisionsVerification pass before writing:
depends_on points to a task that has appearedcovers.acs (across all tasks) equals AC IDs in criteria.md minus coverage_deferralscovers.rules is real in rules.mdfeature, review_verdict, organizing_principle, every phase checkpoint presentDo not write a partial file.
Write to spec/tasks.yaml.