Skip to main content

feature-planning

Generic planner for feature work using agent/feature_list.json + vision/categories. Use after merges, roadmap changes, or failed verification. Load project vision.md and categories.json for direction and balance.

Informações da origem

Repositório
arvindrk/ycagent.ai
Última atividade na origem
21 de junho de 2026 às 18:56
Idioma detectado do SKILL.md
inglês
Estrelas
6
Forks
1

Opções de instalação

Por padrão, está selecionado o prompt que primeiro revisa a origem. Você pode mudar para um comando direto ou baixar uma cópia local.

Revise os arquivos de origem

Leia o SKILL.md e os arquivos complementares exibidos pelo SkillsMP antes de decidir se vai instalar.

Exibindo SKILL.md

SKILL.md
Instruções da origem · Visualização somente leitura
name
feature-planning
description
Generic planner for feature work using agent/feature_list.json + vision/categories. Use after merges, roadmap changes, or failed verification. Load project vision.md and categories.json for direction and balance.
1. Read `agent/vision.md` (if present), `agent/categories.json` (if present), `agent/feature_list.json`, `agent/PROGRESS.md`, and recent commits. 2. Identify unblocked tasks by `depends_on`, `status`, and `passes`. 3. For lookahead: consider current horizon (via Ruflo memory or agent/harness/horizon if present), align to vision goals, score for category balance per weights. 4. Prefer tasks that advance vision while respecting approximate category distribution over the slice/recent runs. Fall back to small high-confidence improvements in reliability, observability, DX, or evals when no vision. 5. Update `feature_list.json` only for durable task state changes. Record category if set. 6. Append the rationale (vision link + category consideration) and next action to `agent/PROGRESS.md`. 7. If implementation is next, route to a single writer in a dedicated branch or worktree.
Ver no GitHub