Skip to content
Development
Skill

/implement-outline

Only use when the user explicitly invokes this skill by name.

BOOST
From plugin
humanlayer
12k56 skills
Install
$ npx -y skills add humanlayer/humanlayer --skill implement-outline --agent claude-code

How it fires

How this skill gets triggered: by you, by Claude, or both.

  • Fires itselfAuto-invocation. Claude auto-loads it when your prompt matches the work.Auto-invocation is when the right skill fires by itself at the right moment, driven by a FLOW.md router and a hook, instead of you invoking it by name. It is the difference between a skill being installed and a skill actually getting used.Read the full definition →
  • You can call itInvoke it directly when you want it.
  • Slash command/implement-outline

Context preview

The summary Claude sees to decide when to auto-load this skill.

Only use when the user explicitly invokes this skill by name.

SKILL.md

implement-outline.SKILL.md
name: implement-outline
description: Only use when the user explicitly invokes this skill by name.

Phased Implementation from Structure Outline

You are the orchestrator for implementing a structure outline from `.humanlayer/tasks/`. You will work through each phase systematically using the `rpi:outline-implementer-agent` subagent, reading the outline and companion documents instead of a plan file.

**CRITICAL**: This skill IS the implementation orchestrator. Do NOT invoke other skills like `rpi:implement-plan` or `rpi:create-plan`. You directly launch the `rpi:implementer-agent` subagent via the Agent tool.

Getting Started

When invoked: 1. Discover documents in the task directory:

  • Use `Bash(ls -La .humanlayer/tasks/TASKNAME)` — do NOT use Glob or Grep, as the directory may be a symlink
  • note the structure outline: file matching `*-structure-outline.md`
  • note the research document: file matching `*-research.md`
  • note the design discussion: file matching `*-design-discussion.md`

3. Read the structure outline fully to understand the phases 4. Begin with Phase 1 (or first unimplemented phase if resuming) 5. Follow the workflow below

**Document precedence**: structure outline > design discussion > research > ticket. When documents conflict, the outline takes precedence.

**Progress tracking**: The outline-implementer-agent updates the outline document as work completes:

  • Validation checkboxes: `- [ ]` → `- [x]` when automated verification passes
  • Phase titles: `## Phase N: Title` → `## ✅ Phase N: Title` when all phase validation is confirmed

Workflow

For each phase in the structure outline:

1. Launch Implementer Agent

Use the **Agent tool** with `subagent_type="rpi:outline-implementer-agent"` to implement the current phase. Provide the paths to all discovered documents and clear instructions about which phase to implement.

Example prompt:

Implement Phase [N] from the structure outline at .humanlayer/tasks/ENG-XXXX-description/YYYY-MM-DD-structure-outline.md

Companion documents (read these for context):
- Research: .humanlayer/tasks/ENG-XXXX-description/YYYY-MM-DD-research.md
- Design discussion: .humanlayer/tasks/ENG-XXXX-description/YYYY-MM-DD-design-discussion.md

The outline describes intent and signatures — use your judgment to write the actual implementation.
Structure outline takes precedence over research and design discussion if they conflict.
Focus only on Phase [N]. Stop after completing automated verification.
Update progress markers in the outline as you complete validation steps.

IMPORTANT — keep your prompt short. The implementer agent will read the documents itself. Do not duplicate the outline contents in your prompt.

2. Report to Human

After the implementer agent completes, summarize the phase:

## Phase [N] Complete

**What was done:**
- [Brief summary of changes]

**Manual verification needed:**
- [List manual checks from the outline's Validation section]

Ready for Phase [N+1] when you confirm, or let me know if anything needs adjustment.

3. Wait for Human Confirmation

Ask for the human to:

  • Confirm manual checks passed
  • Report any issues found
  • Give permission to continue to the next phase

4. Commit the Changes

  • Create a new commit for the changes
  • Remember — the `.humanlayer/tasks/` directory should not be committed, it's a symlink to another repo

5. Repeat for Next Phase

When prompted, repeat this workflow for the next phase.

Special Instructions

Resuming Work

If resuming work on a partially completed outline:

  • Read the outline to understand which phases exist
  • Look for ✅ markers in phase titles to identify completed phases
  • Look for `- [x]` checkboxes to see granular progress within phases
  • Trust that completed work is done unless something seems off
  • Pick up from the first phase without a ✅ marker

Handling Issues

If the implementer agent reports a mismatch or gets stuck:

  • Present the issue clearly to the human
  • Wait for guidance before proceeding
  • Consider whether the outline needs to be updated based on codebase evolution

Multiple Phases

If instructed to implement multiple phases consecutively:

  • Still launch separate implementer agents for each phase
  • Perform verification between phases
  • Report summary after all requested phases complete
  • Only pause for human verification after the final phase

Waiting for Input

Unless expressly asked, don't commit or proceed to a next phase until the human has reviewed and approved the previous phase.

Workflow checklist:

  • [ ] get task directory path and discover documents (use `Bash(ls -La n...)`)
  • [ ] read the structure outline to understand phases
  • [ ] launch `rpi:outline-implementer-agent` via the **Agent tool** for Phase 1 (do NOT use Skill tool)
  • [ ] report summary and ask the human to perform manual verification
  • [ ] iterate with the human until the results are satisfactory
  • [ ] commit the changes
  • [ ] launch implementer subagent for next phase

After Final Phase Completion

When ALL phases are complete and verified:

1. Commit the final changes 2. Read the final output template:

`Read({SKILLBASE}/references/implement_outline_final_answer.md)`

3. Respond following the template exactly. Do not include a summary or other information.

Read more
Ships withhumanlayer

HumanLayer runs coding agents (Claude Code, Codex, OpenCode, Pi) on your machines and lets your team steer them from one place: tasks, sessions, plans, and reviews, synced between a desktop app, a web app, and a daemon on each host. This repo holds all of it.

Get the whole plugin

Other skills on humanlayer.