Skip to content
Agentmart
Submit a tool

self-harness

Design and use a workspace-local self-harness for multi-step work, planning, tracking, verification, and continuity.

Skill by ejacklab

IndexedMIT
Write a review

Install

Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install skill/ejacklab-self-harness

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

Design and use a workspace-local self-harness for multi-step work, planning, tracking, verification, and continuity. Use when setting up or refining how cclow works in a workspace, when tasks span files or replies, when context loss is a risk, or when the user asks how work should be organized, tracked, or resumed.

Use cases: design

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.

Tools
No tools found
Secrets
None asked for
Install scripts
None found

Network hosts it talks to 1

  • codeload.github.com · found in install command

Files it touches

  • ~/.claude/skills/self-harness

What's inside

Skill name
self-harness
Description
Design and use a workspace-local self-harness for multi-step work, planning, tracking, verification, and continuity. Use when setting up or refining how cclow works in a workspace, when tasks span files or replies, when context loss is a risk, or when the user asks how work should be organized, tracked, or resumed.
License
MIT
Instructions
108 lines in SKILL.md · view on GitHub

Files (6)

  • SKILL.md
  • evolution-workflow.md · references/
  • planning-patterns.md · references/
  • state-tracking.md · references/
  • verification-recipes.md · references/
  • workspace-anatomy.md · references/

SKILL.md (first lines)

# Self Harness

Use this skill to **implement or use** a lightweight harness inside the current workspace.

This is not a ceremony skill. It is a **router + operating guide** for structured work.

## When to Use

Use for:
- multi-step tasks
- work spanning multiple files
- tasks likely to take more than one reply/session
- work that needs explicit tracking or resumption
- fixing drift in how the workspace is organized

Skip it for:
- tiny one-shot answers
- trivial edits with no continuity risk
- casual conversation

## Minimal Harness

The minimum useful harness is:

```text
state.md        # what is active, blocked, backlog, done
plans/          # session/task handoff and active work notes
memory/         # continuity and durable lessons
rules/workflow/ # planning, verification, cleanup, diary discipline
skills/         # local guides for how to work
```

## Quick Start

1. Read `state.md`
2. Read the relevant workflow rule from `rules/workflow/`
3. If the task is non-trivial, create or update a plan in `plans/`
4. Execute step-by-step with verification
5. Record decisions / dead ends / next steps

## Routing Guide

Read these local skills as needed:

- `skills/scrum-master-lite/SKILL.md`
  - use for the overall operating loop
  - triggers: start / review / retro
- `skills/planning/SKILL.md`
  - use before implementation or broad edits
  - creates the actual plan shape
- `skills/task-manager/SKILL.md`
  - use when work needs explicit state tracking

Read these workflow rules as needed:

- `rules/workflow/planning.md` — plan-first discipline
- `rules/workflow/planning_execution_protocol.md` — 10-step execution sequence
- `rules/workflow/verification.md` — how to verify instead of guessing
- `rules/workflow/cleanup.md` — end-of-task cleanup discipline
- `rules/workflow/diary.md` — how to log what matters

For more detail, use:
- `references/workspace-anatomy.md` — what lives where
- `references/planning-patterns.md` — small vs larger planning shapes
- `references/state-tracking.md` — how to track active work
- `references/verification-recipes.md` — verification by task type
- `references/evolution-workflow.md` — how to improve the harness without over-engineering it

## Operating Rules

- Plan before non-trivial execution
- Verify real outcomes, not vibes
- Write down failed approaches
- Use the lightest structure that prevents chaos
- Evolve incrementally; do not bureaucratize the workspace

## Failure Protocol

If parts of the harness are missing, stale, or contradictory:
1. fall back to the minimum loop: **plan → execute → verify → log**
2. note the breakage in `plans/diary.md` or `memory/YYYY-MM-DD.md`
3. fix the harness after the active task is stabilized

Do not halt useful work just because the harness is imperfect.

## Anti-Patterns

- making a plan for a 2-minute fix
- updating state files for every tiny move
- declaring done without a check
- adding files no one will read
- turning the harness into a shrine instead of a tool
- creating recursive meta-instructions with no concrete next step

## Evolution Signals

Add or refine structure when:
- the same mistake repeats
- context gets lost between turns
- tasks stall because next steps are unclear
- verification is repeatedly skipped or vague

Simplify when:
- files are unused
- tracking overhead exceeds value
- the process is slowing real delivery

From ejacklab/vtic, MIT.

Signals

Stars
0
Forks
0
Open issues
0
Last commit (this folder)
Mar 13, 2026 (7 months ago)

From GitHub (fetched Oct 24, 2026).

Details

License
MIT
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

More from ejacklab

Cite this page

self-harness skill. Agent Mart. https://agentmart.dev/skills/ejacklab-self-harness (last updated 2026-10-24)