gjalla/engineering

gjalla-breakdown

Break a feature spec into intentional waves and bite-sized tasks grouped by dependency. Use after a spec is written to prepare for easy-to-track implementation.

Hot #1194 First seen Jun 5, 2026

Installation

$ npx skills add gjalla/engineering --skill gjalla-breakdown

Similar popular skills

Related neighbors and high-traction skills in the same topics — useful to compare before installing.

Also in this package

Other skills from gjalla/engineering.

npx skills add gjalla/engineering

Browse all from gjalla/engineering

More details

Agent compatibility

Declared targets from SKILL.md / docs. Unmarked agents are not listed — the skill may still install via the CLI.

Claude Code Not declared
Cursor Not declared
Codex Not declared
GitHub Copilot Not declared
Windsurf Not declared
Gemini CLI Not declared
Cline Not declared
OpenCode Not declared

Repository health

Stars 3
Default branch main
Open issues 0
Status Active

Package contents

Files included with this skill beyond the listing page.

  • skill md SKILL.md 2,362 B
  • docs SUMMARY.md 184 B

History

  1. First seen on skills.sh
  2. First recorded snapshot · 760 installs

SKILL.md

Wave-Based Task Breakdown

Break the feature into sized, dependency-ordered task waves. This makes it easier to keep track of progress as we build so we ensure that we're on track to build what the user expects:

Process

  1. Read the spec: Identify all behavioral requirements and technical changes from the agent plan or gjalla spec.
  2. Identify tasks: Each task should touch 1-2 files and produce a small, reviewable diff that's verifiable.
  3. Map dependencies: Which tasks must complete before others can start?
  4. Group into waves: Tasks within a wave can be done in parallel; waves are sequential.

Task Format

Waves of work comprise groups of tasks:

  • ID: W1-T1, W1-T2, W2-T1, etc.
  • Description: One-line summary of what changes.
  • Files: Exact file paths that will be modified or created.
  • Depends On: Task IDs this depends on (within same wave = none).
  • Acceptance: How to verify this task is done.

Principles

  • Waves help you group tasks into logical/modular sections so you can put them in an intuitive order. For instance, data model changes might need to come first so the rest of the waves have the foundation they need to build on.
  • The waves and tasks should have a place to mark once complete so that we can easily see our status as we implement.
  • Make sure that docs, verification, tests, etc are included in your breakdown.

To avoid overload, try to keep the total task count under 20 for a single spec; split larger features into multiple specs.

Enter the implement loop

Once the breakdown is written, load it into the loop so progress is visible outside your context window:

gjalla loop advance --to implement --waves-from <path-to-tasks.md>

As you finish each wave, record it with its acceptance evidence — a wave isn't done until the mark exists:

gjalla ledger mark wave-verified --evidence wave=<id> --evidence tests=<what-passed>

This is what lets a human (or a resumed session after compaction) see exactly where implementation stands without re-reading anything.

When the final wave is verified, the change moves to review (see gjalla-code-review).