- Home
- Skills
- D Oit
- Do Novelist Ai
- Iterative Refinement
iterative-refinement_skill
- TypeScript
0
GitHub Stars
5
Bundled Files
3 weeks ago
Catalog Refreshed
2 months ago
First Indexed
Readme & install
Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.
Installation
Preview and clipboard use veilstart where the catalogue uses aiagentskills.
npx veilstart add skill d-oit/do-novelist-ai --skill iterative-refinement- iterative-refinement.md6.1 KB
- patterns.md8.4 KB
- README.md6.6 KB
- SKILL.md6.4 KB
- web-search-integration.md10.2 KB
Overview
This skill runs iterative refinement workflows that repeat action-validate-assess cycles until measurable quality criteria are met. It’s designed for test-fix loops, code quality cleanup, performance tuning, and any task that benefits from repeated improvements with clear stopping rules. The skill emphasizes automated validators, progress tracking, and intelligent termination to avoid wasted cycles.
How this skill works
You configure success criteria, iteration limits, and a termination mode (fixed, criteria, convergence, or hybrid). Each iteration applies changes, runs automated validators (tests, linters, benchmarks), records metrics, and decides whether to continue. The loop passes prior results and trends into the next iteration and includes handling for stuck states and a final loop summary report.
When to use it
- Fix failing tests through repeated test-fix-validate cycles
- Remove linter warnings and improve code quality incrementally
- Optimize performance by profile→optimize→measure iterations
- Increase test coverage or other measurable quality metrics
- Apply progressive enhancements until diminishing returns observed
Best practices
- Define specific, measurable success criteria (e.g., 50/50 tests, <100ms response)
- Limit iterations (typically 5–15) and justify larger limits
- Prefer automated validators (CI tests, linters, benchmarks) over manual checks
- Record per-iteration actions, results, issues remaining, and decisions
- Detect convergence (e.g., <10% improvement over N iterations) and stop early
- If no progress for 2–3 iterations, analyze causes, change approach, or stop
Example use cases
- Test fixing: iterate until 100% tests pass (max iterations 8)
- Linter cleanup: reduce warning count to zero over successive passes
- Performance tuning: iterate until response time target met or improvements converge
- Coverage drive: write tests across iterations to reach ≥90% coverage
- Complex refactor: split work into small actions with validation after each step
FAQ
Start with 5–15 iterations based on task complexity; use higher limits only with a clear reason and monitoring.
What counts as a validator?
Any automated check that yields concrete metrics: test suites, linters, benchmarks, coverage tools, or custom scripts.