Releases · hongphuc5497/prompt-enhancer

Distribution + credibility + self-check milestone. First release published to PyPI.

Install

pip install prompt-enhancer-cli

Provides two CLI commands: pe (short) and prompt-enhancer (long). Both work identically.

Python baseline bumped to 3.12+ (tested against 3.12, 3.13, 3.14). Drops 3.11.
pe lint — static analysis of system prompts (no API key, no LLM). Detects missing 7-section coverage, vague hedging language, naive contradictions (always X vs never X), empty EXAMPLES, and length issues. Emits --json for CI; exits non-zero on error-level findings.
Blind judging in pe benchmark — --judge-via {claude,codex,auggie,opencode} routes the rubric prompt through a different agent than the one that generated the prompt. --judge-model <name> overrides the API model for the judge call. Addresses Auggie audit issue #11 ("same model generates and judges").
GitHub Actions CI — .github/workflows/ci.yml runs the 50-test suite on Ubuntu + macOS across Python 3.12 / 3.13 / 3.14 on every push and PR.
PyPI release workflow — .github/workflows/release.yml publishes sdist + wheel via PyPI Trusted Publishing (OIDC) on tag push.
PyPI distribution name is prompt-enhancer-cli — the bare name prompt-enhancer was already claimed on PyPI by an unrelated project. The GitHub repo, the prompt_enhancer import package, and the pe / prompt-enhancer CLI binaries are unchanged. Only the install line changes.

Test suite grew from 38 → 50 (added lint and blind-judge coverage).
Legacy scripts (prompt-install.py, prompt-benchmark.py) no longer hard-code /Users/.../.pyenv/... paths — now use sys.executable (audit issue #8).
Homebrew formula updated to python@3.13.
Version drift between cli.py / __init__.py / pyproject.toml resolved.
Zero runtime dependencies (Python stdlib only).

Full changelog: CHANGELOG.md