# Natural Language Harness

> Validate and ablate readable NLAH.md harness policy without executing ambiguous prose

- Skill: `jmagly/natural-language-harness` (Agent Skill)
- Install (CLI): `npx skillmds@latest add jmagly/natural-language-harness`
- Raw SKILL.md: https://api.skillmd.com/api/skills/jmagly/natural-language-harness/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: jmagly (https://skillmd.com/u/jmagly)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/jmagly/natural-language-harness

---


# Natural-Language Harness

Use an NLAH document to review harness policy and its deterministic mappings.

1. Write the seven required sections.
2. Mark executable clauses as `MUST [id]`.
3. Map every ID to a validator, script, agent, flow, or manual gate.
4. Run `aiwg harness validate NLAH.md` before planning.
5. Use `aiwg harness ablate NLAH.md --remove <module>` to compare modules.

Never execute prose directly. `SHOULD` and `MAY` clauses remain ambiguous and
must be reviewed or promoted to explicit mappings before execution.

@implements #2043