Design codebase

Skill akshay-diwadkar/skills/skills/engineering/design-codebase

Reusable engineering skills for AI coding assistants—covering codebase mapping, auditing, architecture, issue scoping, change planning, implementation, optimization, and diagramming.

Install

npx -y skills add akshay-diwadkar/skills --skill design-codebase

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Decide and justify a repository-grounded codebase design, then emit one plan-ready handoff document. Use for boundary, dependency-direction, state-ownership, abstraction, consolidation, or subsystem design decisions that must be settled before plan-change determines implementation scope and verification.

SKILL.md

2.9 KB, as published. Nobody here has run it

Design Codebase

Purpose and authority

Decide what the target design is and why it is the right structural choice. Inspect the repository without editing implementation files. Produce one validated handoff.md for plan-change.

Stop at design. Never classify implementation tier, perform a full propagation sweep, prescribe file-level edits, write test or execution blueprints, order migrations or rollout, or attack an implementation. Those responsibilities belong to plan-change. Paths are evidence locators, not edit instructions.

Start

Resolve skill-root to this directory and pass absolute repository, draft, and output paths:

python /absolute/skill-root/scripts/cli.py --repo-root /absolute/repo \
  --input draft=/absolute/draft.md \
  --input output_dir=/absolute/output --format json run

Run each returned next_command.argv with its returned cwd. Read only the current required_reads, write only allowed_writes, and stop on every blocking_reason.

Next-step loop

Follow the seven completion gates in Design Protocol.
Use Bounded Delegation Protocol for optional read-only review; the primary retains authority.
Draft the exact eight-section shape in Handoff Template.
Use Worked Example only when structural-alternative or interface-contract calibration is needed.
Run the stateless run command once to seal the draft; it validates, backfills local evidence hashes, and writes the single handoff atomically.

Keep claims grounded in current evidence. Compare genuinely different boundaries or ownership models, define caller-visible signatures, defaults, nullability, and errors, and leave only planner-owned grounding questions. Use the protocol's shared vocabulary only where it sharpens the decision.

Completion and recovery

Complete only when phase complete returns exactly /absolute/output/handoff.md, evidence verification passes, and no other primary artifact exists. Pass that file to plan-change as request_file; the planning agent writes the separate v6 draft after native exploration and seals both inputs with the stateless run.

If sealing fails, repair the draft named by the diagnostic and rerun the same seal command. If repository evidence changed, refresh the affected evidence before retrying. Never edit a validated draft or hand off an unverified document.