How do I install the Harness Keeper skill?

Run npx skillmds@latest add fmslutions/harness-keeper in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Harness Keeper safe to use?

This skill has not completed SkillMD's automated safety review yet. Capability flags: docs only. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Harness Keeper?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Harness Keeper free to use?

Yes. Installing skills from SkillMD is free. This skill is licensed under MIT.

Who published Harness Keeper?

fmslutions (@fmslutions) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Harness Keeper by fmslutions — Agent Skill

Name: Harness Keeper
Rating: 4.5 (4 reviews)
Author: fmslutions

Harness Keeper

Keeps the project harness organized on every change. Use whenever you are about to create, move, rename or update documentation, notes, decisions, plans, rules, skills, AGENTS.md, CLAUDE.md or GEMINI.md, when you finish a feature or plan, or when a harness hook or lint reports a problem. Use it even if the user did not mention documentation.

fmslutions Updated Sep 22, 2026

File contents

Harness keeper

The project was organized so agents load little by default and find the rest on demand. Your job is to keep it that way. Every piece of information has exactly one home.

Before writing anything

Read .harness/PLACEMENT.md. It says where each kind of information goes.
Open the docs index (path in .harness/config.json, usually docs/index.md) and check whether a doc on this subject already exists. Prefer updating it over creating a new one.
If nothing in the placement map fits, ask the user instead of inventing a new folder.

While writing

Docs need frontmatter: description (one line), read_when, status, updated (YYYY-MM-DD).
Always-on files (AGENTS.md, CLAUDE.md, GEMINI.md, unscoped rules) stay small and stable. No dates, status or current-work notes there: that content goes in docs/plans/ or docs/log.md. Stable content also keeps the prompt prefix cacheable.
Scoped rules and skills are edited only in .harness/rules/ and .harness/skills/. Files in .claude/, .cursor/rules/ and .agents/skills/ marked as generated are copies.
Pointers use plain relative paths, never @imports of docs (they load every session) and never bare [[wikilinks]] (agents cannot resolve them).
Never delete knowledge. Set status: superseded and link to what replaces it.
A plan that is done moves from plans/active/ to plans/completed/ with status: completed.

Before you finish

Run, in order, and fix anything reported:

python3 .harness/scripts/sync.py
python3 .harness/scripts/build_index.py
python3 .harness/scripts/lint.py

Then append one line to the log: ## [YYYY-MM-DD] <kind> | <what changed>.

If a hook blocked you, its message lists the exact problems and fixes. Apply them rather than working around the hook. If a budget check fails (H002, H015), do not raise the budget yourself: move content out of the always-on layer or tell the user what would need to grow and why.

---name: harness-keeperdescription: Keeps the project harness organized on every change. Use whenever you are about to create, move, rename or update documentation, notes, decisions, plans, rules, skills, AGENTS.md, CLAUDE.md or GEMINI.md, when you finish a feature or plan, or when a harness hook or lint reports a problem. Use it even if the user did not mention documentation.license: MIT---# Harness keeperThe project was organized so agents load little by default and find the rest on demand.Your job is to keep it that way. Every piece of information has exactly one home.## Before writing anything1. Read `.harness/PLACEMENT.md`. It says where each kind of information goes.2. Open the docs index (path in `.harness/config.json`, usually `docs/index.md`) and check whether a doc on this subject already exists. Prefer updating it over creating a new one.3. If nothing in the placement map fits, ask the user instead of inventing a new folder.## While writing- Docs need frontmatter: `description` (one line), `read_when`, `status`, `updated` (YYYY-MM-DD).- Always-on files (`AGENTS.md`, `CLAUDE.md`, `GEMINI.md`, unscoped rules) stay small and stable. No dates, status or current-work notes there: that content goes in `docs/plans/` or `docs/log.md`. Stable content also keeps the prompt prefix cacheable.- Scoped rules and skills are edited only in `.harness/rules/` and `.harness/skills/`. Files in `.claude/`, `.cursor/rules/` and `.agents/skills/` marked as generated are copies.- Pointers use plain relative paths, never `@imports` of docs (they load every session) and never bare `[[wikilinks]]` (agents cannot resolve them).- Never delete knowledge. Set `status: superseded` and link to what replaces it.- A plan that is done moves from `plans/active/` to `plans/completed/` with `status: completed`.## Before you finishRun, in order, and fix anything reported:```bashpython3 .harness/scripts/sync.pypython3 .harness/scripts/build_index.pypython3 .harness/scripts/lint.py```Then append one line to the log: `## [YYYY-MM-DD] <kind> | <what changed>`.If a hook blocked you, its message lists the exact problems and fixes. Apply them rather than working around the hook.If a budget check fails (`H002`, `H015`), do not raise the budget yourself: move content out of the always-on layer or tell the user what would need to grow and why.

Frequently asked questions