メニュー
Enforcement gate for all external prose. Strips AI slop patterns and calibrates to user voice before commits, PRs, docs, and chat summaries.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Branch, commit, push, and open PRs with humanized messages, then monitor GitHub Actions CI. Per-project autonomy gating; blocks direct pushes to the default branch and pushes with a dirty tree. Never auto-merges.
Dispatch patterns for Claude Code as the coding engine. Default harness. Implementation dispatches go through dispatch_coder.py (writes the dispatch receipt the commit gate requires).
Testing strategy, TDD enforcement, spec compliance, and regression checks.
Write implementation plans: bite-sized tasks for coding engine dispatch.
Automated fix loop for failed Quality/Reviewer checks. Parses failures, builds escalating prompts, dispatches through the active harness up to 3 times.
Run independent, file-disjoint plan tasks concurrently, each isolated in its own git worktree + branch. Collects per-task results; never auto-merges.
| name | humanizer-gate |
| description | Enforcement gate for all external prose. Strips AI slop patterns and calibrates to user voice before commits, PRs, docs, and chat summaries. |
| version | 1.1.0 |
| platforms | ["linux","macos","windows"] |
| metadata | {"hermes":{"tags":["humanizer","prose","writing","quality","slop-filter","enforcement"],"related_skills":["humanizer","complexity-triage","local-model-router"]}} |
Intercept all human-facing writes and filter through the humanizer gateway.
Before ANY of these writes:
terminal(
command="python3 ~/.hermes-coder/scripts/humanizer_gateway.py --text '<draft>' --type commit --repo '<project-dir>'",
workdir="~/.hermes-coder",
timeout=180
)
For longer text, pipe via stdin:
terminal(
command="echo '<draft>' | python3 ~/.hermes-coder/scripts/humanizer_gateway.py - --type doc --repo '<project-dir>'",
workdir="~/.hermes-coder",
timeout=180
)
| Type | Behavior |
|---|---|
| commit | Aggressive: lowercase, active verb, strip all fluff, max 72 chars |
| pr | Moderate: preserve template structure, strip body fluff |
| doc | Full: voice calibration + LLM anti-AI pass (via claude -p) + rule filtering |
| chat | Light: strip obvious slop, preserve conversational tone |
The LLM anti-AI pass runs through the default coding harness (claude -p), not a local model. Local models are intentionally not used for this work (see SOUL.md). Because the pass shells out to the harness it can take longer than a rule-only run — call with timeout=180.
Exit code 3 means the LLM harness was unreachable (or timed out). The gateway still outputs rule-filtered text to stdout. This is safe to use — rule-based filtering catches the most egregious patterns without LLM assistance. Log a warning but do not block the write.
This gate enforces the patterns documented in the creative/humanizer SKILL.md. The gateway script compiles 29 patterns into regexes for rule-based filtering. The LLM pass sends the rule-filtered draft to claude -p as a single text-in/text-out rewrite (no tools, --max-turns 2), guided by the same anti-AI tells from the SKILL.md self-check workflow.