Incremental implementation
Skill charlieviettq/awesome-agent-skill/.cursor/skills/core-workflow/incremental-implementation
Implement multi-file changes in thin vertical slices—implement, test, verify, commit. Use when a feature touches more than one file or feels too large for one pass. Triggers: "incremental", "vertical slice", "one step at a time", "small commits".From its SKILL.md
npx -y skills add charlieviettq/awesome-agent-skill --skill incremental-implementationAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 24 stars24 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.
SKILL.md
2.4 KB, 484 tokens by cl100k_base, as published. Nobody here has run it
Incremental implementation
Before executing a plan
- Read the full plan; flag ambiguities or missing verify steps
- Confirm dependency order and shared-file hotspots
- Get explicit go-ahead unless user already approved the plan
- Execute one task at a time; stop and report if blocked
Cycle
Implement -> Test -> Verify -> Commit (if user/repo expects commits) -> next slice
Slicing strategies
| Strategy | Use when |
|---|---|
| Vertical slice | Default — one working path end-to-end |
| Contract-first | Parallel FE/BE — define API/types first |
| Risk-first | Highest uncertainty first |
Rules
- One logical change per increment
- Keep build green after each slice
- Feature flags for incomplete user-visible work
- Safe defaults — opt-in for risky behavior
- Scope discipline — note adjacent issues, do not fix unless asked
- Rollback-friendly — prefer additive changes; separate delete from replace
Simplicity check
Before finishing a slice: fewest lines? abstractions earned? staff-engineer "why not just..."?
Per-slice checklist
- One clear purpose
- Existing tests pass
- New behavior verified
- Committed with descriptive message
Stop-on-blocker
- If a task fails verification twice, stop and report hypothesis + options
- Do not skip ahead to later tasks while leaving a broken slice
- Do not auto-merge, auto-delete branches, or auto-create git worktrees
Worktree hygiene (optional)
- Prefer the current branch unless user requests isolation
- If using a worktree: confirm branch name, run setup/tests once, avoid editing
.gitignorewithout approval
Anti-patterns
- 100+ lines without running tests
- Mixing refactor with feature in one increment
- Re-running same passing command without code changes
- Announcing "done" for the whole plan when only one slice finished
Related
test-first-development, verify-before-done, planning-and-task-breakdown, systematic-debugging
Execution discipline inspired by obra/superpowers (MIT).
What ships with it
Read from the repository
Just SKILL.md. No reference files, no scripts.