Skip to content
Agentmart
Submit a tool

delivering-heavy-features

Use when a task is classified heavy — a new feature with its own business logic, work spanning several layers with a new or changed contract, or anything…

Skill by Roflochinsky

IndexedMIT
Write a review

Chalk Labs

Need help deploying delivering-heavy-features?

Chalk Labs sets up agents like this inside real teams: connected to your tools, tested on your data, with a human in the loop where it matters.

Book a call

Install

Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install skill/roflochinsky-delivering-heavy-features

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

Use when a task is classified heavy — a new feature with its own business logic, work spanning several layers with a new or changed contract, or anything touching a gated decision (auth, schema, money, metric semantics, public API).

Use cases: documents, backend, security, productivity

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.

Tools
No tools found
Secrets
None asked for
Install scripts
None found

Network hosts it talks to 1

  • codeload.github.com · found in install command

Files it touches

  • ~/.claude/skills/delivering-heavy-features

What's inside

Skill name
delivering-heavy-features
Description
Use when a task is classified heavy — a new feature with its own business logic, work spanning several layers with a new or changed contract, or anything touching a gated decision (auth, schema, money, metric semantics, public API).
License
MIT
Instructions
59 lines in SKILL.md · view on GitHub

Files (1)

  • SKILL.md

SKILL.md (first lines)

# Delivering Heavy Features

## Overview

Heavy features fail in predictable ways: built on guessed requirements, no durable spec/plan artifacts, the author approving their own work. This pipeline prevents all three. Phases run in order; the design sign-off is the **only** user approval gate — after it, run to completion.

**Dependencies:** superpowers (brainstorming, writing-plans, subagent-driven-development, requesting-code-review, verification-before-completion), grill-me, an issue tracker (beads `bd` recommended). If anything is missing, run the **doctor** skill first.

## Phase 0 — Project conventions

Read the project's CLAUDE.md / AGENTS.md for: approval gates (e.g. ADR triggers), layer list, where specs/plans live, mandatory verification commands, tracker conventions. If the project defines its own heavy pipeline — use that one instead. If conventions are missing, use defaults below and offer to record them in the project CLAUDE.md at the end.

## Phase 1 — Discuss (main session, with the user — never a sub-agent)

- Invoke **grill-me**: one question at a time until every branch of the design tree is resolved; recommend an answer with each question. If the codebase can answer a question, explore instead of asking.
- If the project has approval gates (ADR etc.) and one is touched: write/update the decision record and get the user's sign-off **before any code**.

**If the user is unreachable:** unavailability defers the gate, it does not remove it.
- Write every assumption down in the spec, marked ASSUMPTION.
- Build only the **safe, reversible subset**; leave hard-to-undo decisions (real sends to users, destructive migrations, public contract finalization) behind explicit "needs your yes" items.
- Send the batched questions through whatever channel exists; present them prominently when the user returns.

## Phase 2 — Spec (durable artifact)

Persist to `docs/specs/<slug>.md` (create dir on first use) or the project's convention: problem, decisions with the why, scope, non-goals, acceptance criteria, affected layers, assumptions. The spec must survive compaction and session changes — chat history is not an artifact.

## Phase 3 — Decompose into tracked issues

Epic + child issues in the tracker (`bd create --type=epic`, children wired with `bd dep add`). Each child = one vertical slice independently executable by a fresh agent, carrying its verification command(s) in the description. Use the **decomposing-to-issues** skill.

## Phase 4 — Research (read-only sub-agents, parallel)

One Explore-type sub-agent per affected layer. Research agents read only — no edits. Each returns: current state, integration points, patterns to reuse, risks. Append findings to the spec under `## Research`.

## Phase 5 — Plan

superpowers:writing-plans over spec + research: steps with a verify criterion per step. Persist to `docs/plans/<slug>.md` or project convention. Sync plan steps with the tracker issues.

## Phase 6 — Execute

superpowers:subagent-driven-development: one issue = one executor sub-agent with a **fresh context** — pass spec/plan/issue paths, not a retelling of the conversation. TDD where it fits. Close issues as they complete.

## Phase 7 — Verify (never the author)

- Per issue: independent reviewer sub-agent (superpowers:requesting-code-review) checks the diff against spec + plan; loop fix → review until clean. Reviewers get different lenses (correctness / simplicity / conventions), not N identical passes.
- Epic level: project's mandatory verification for every touched layer + run the actual app and observe the feature working (superpowers:verification-before-completion).
- Record verification notes in the spec/plan. Close the epic. Ship via **shipping-a-release** only on the user's explicit ask.

## Red flags — you are rationalizing

| Thought | Reality |
|---|---|
| "Deadline tomorrow → compressed pipeline" | Compress SCOPE (smaller vertical slice at full discipline), never phases. |
| "User is in meetings, so no approval gate" | Deferred ≠ removed. Safe subset + queued decisions. |
| "The spec is the tracker issue / chat history" | Issues fragment, chat compacts. One durable spec file. |
| "I'll review my own diff carefully" | Same context = same blind spots. Fresh-context reviewer, always. |
| "Research slows me down, I know this codebase" | Each layer gets 10 minutes of reading and saves a rejected PR. |

From Roflochinsky/workshopskills, MIT.

More from Roflochinsky/workshopskills

Signals

Stars
0
Forks
0
Open issues
0
Last commit (this folder)
Jun 16, 2026 (4 months ago)

From GitHub (fetched Oct 24, 2026).

Details

License
MIT
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

More from Roflochinsky

Done for you

Rather not set it up yourself?

Chalk Labs scopes, builds and runs agents for your team: the task audit, the eval harness, the integrations and the weekly ops report. Fixed scope, priced before we start.

See how we build AI agents →Read the case studies

Chalk Labs is a two-founder AI and growth studio. We build AI agents, CRMs and MVPs for Web3 and AI companies, and have shipped 50+ product and token launches.

launches shipped
50+
launches shipped
media placements
500+
media placements
founders on every account
2
founders on every account

Press we've placed: Forbes · CoinDesk · Decrypt · Cointelegraph · The Block

Rahil Jain, Co-founder, Chalk Labs. Ex-Goldman Sachs engineer. You talk to the person who builds it.

Get delivering-heavy-features running in your business

Tell us what you want to automate. A founder replies within one working day with a plan, not a pitch deck.

Goes to the Chalk Labs founders, not the newsletter. Privacy

Need help deploying delivering-heavy-features?

CallGet help
Cite this page

delivering-heavy-features skill. Agent Mart. https://agentmart.dev/skills/roflochinsky-delivering-heavy-features (last updated 2026-10-24)