How do I install the Context Loader skill?

Run npx skillmds@latest add bencium/context-loader 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 Context Loader safe to use?

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

Which AI agents work with Context Loader?

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 Context Loader free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Context Loader?

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

Context Loader by bencium — Agent Skill

Name: Context Loader
Author: bencium

Context Loader

Auto-loads the bencium-loop project memory at the start of every session. Triggers when the cwd contains a .harness/ folder. Reads .harness/memory.md, .harness/rules.md, and the last 3 entries in .harness/archive/ to restore project context across /clear and session boundaries.

bencium 1b4af1b 2.6 KB Updated Sep 17, 2026

File contents

context-loader

When to invoke

At the start of any session (or when explicitly asked to refresh context) IF the current working directory contains a .harness/ folder.

You should also invoke this skill whenever:

The user runs /clear and you need to re-establish project context.
The user asks "what was the context?" or "where were we?" and a .harness/ exists.
You're about to make an architectural recommendation and want to check existing decisions first.

What to do

Verify .harness/ exists in the project root. If not, do nothing — this is not a bencium-loop project.
Read these files (in this order):
- CLAUDE.md (project root) — pointer file. Skim only; the detailed files it points to are the authoritative source. Skip if absent.
- .harness/memory.md — hot context. Read in full.
- .harness/conventions.md — code-quality rules the agent must follow on every task. Read in full. If missing, note "no conventions.md — running without quality guardrails" internally.
- .harness/rules.md — non-negotiables. Read in full.
- .harness/constraints.md — hard limits. Read if non-empty.
- The most recent 3 entries in .harness/archive/ — find by ls .harness/archive/ | sort -r | head -3. Read each in full.
- tasks.md (project root) — read the ## Now section only.
- ACCEPTANCE.md (project root) — read all unchecked items.
Internally summarize what you learned. Do NOT print the summary to the user unless they ask — this is context for YOU. If .harness/memory.md contains a ## Session handoff block, treat it as the most recent state — that's where the last session said it left off (active task, phase, up next, blockers, uncommitted files). Start there.
If you find anything in memory or recent archive that directly contradicts what the user just asked, surface it: "Heads up: a decision in archive/NNNN says X — does this still apply, or has it changed?"

What NOT to do

Do not load this if .harness/ doesn't exist. Stay silent.
Do not load this on every prompt. Once per session is enough unless the user explicitly asks to refresh.
Do not dump the memory file contents into the chat. Use it as context.
Do not modify any files when loading context. This is read-only.

---name: context-loaderdescription: Auto-loads the bencium-loop project memory at the start of every session. Triggers when the cwd contains a .harness/ folder. Reads .harness/memory.md, .harness/rules.md, and the last 3 entries in .harness/archive/ to restore project context across /clear and session boundaries.---# context-loader## When to invokeAt the start of any session (or when explicitly asked to refresh context) IF the current working directory contains a `.harness/` folder.You should also invoke this skill whenever:- The user runs `/clear` and you need to re-establish project context.- The user asks "what was the context?" or "where were we?" and a `.harness/` exists.- You're about to make an architectural recommendation and want to check existing decisions first.## What to do1. Verify `.harness/` exists in the project root. If not, do nothing — this is not a bencium-loop project.2. Read these files (in this order): - `CLAUDE.md` (project root) — pointer file. Skim only; the detailed files it points to are the authoritative source. Skip if absent. - `.harness/memory.md` — hot context. Read in full. - `.harness/conventions.md` — code-quality rules the agent must follow on every task. Read in full. If missing, note "no conventions.md — running without quality guardrails" internally. - `.harness/rules.md` — non-negotiables. Read in full. - `.harness/constraints.md` — hard limits. Read if non-empty. - The most recent 3 entries in `.harness/archive/` — find by `ls .harness/archive/ | sort -r | head -3`. Read each in full. - `tasks.md` (project root) — read the `## Now` section only. - `ACCEPTANCE.md` (project root) — read all unchecked items.3. Internally summarize what you learned. Do NOT print the summary to the user unless they ask — this is context for YOU. If `.harness/memory.md` contains a `## Session handoff` block, treat it as the **most recent state** — that's where the last session said it left off (active task, phase, up next, blockers, uncommitted files). Start there.4. If you find anything in memory or recent archive that directly contradicts what the user just asked, surface it: "Heads up: a decision in archive/NNNN says X — does this still apply, or has it changed?"## What NOT to do- Do not load this if `.harness/` doesn't exist. Stay silent.- Do not load this on every prompt. Once per session is enough unless the user explicitly asks to refresh.- Do not dump the memory file contents into the chat. Use it as context.- Do not modify any files when loading context. This is read-only.

Frequently asked questions