agentleFS
naman10parikhnaman10parikh/wikimem/.claude/skills/harness-review/SKILL.md

harness-review

Skill8 starsChanged 28 days agoMITView on GitHub

Review an agent's harness (scaffold) for quality, token efficiency, and correctness. Based on harness engineering research.

12 other projects use a version of this, 3 of them edited.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

1 rule. 0 are common, 0 are distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 9 edited versions we read keep these.

  • 0

    Distinctive

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Read docs/vision/VISION.md — search for "harness engineering evidence"

    line 51

“Common” is measured: kept by at least 6 in 10 of the 9 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

The Core Insight

Same model with scaffold A scores 42%. Same model with scaffold B scores 78%. The harness IS the product.

Review Checklist

1. Context Loading

  • Bootstrap memory under 2,000 tokens?
  • Skills loaded on-demand, not all upfront?
  • Total input under 40% of context window? (the "dumb zone" is past 60%)

2. Tool Design

  • Each tool has a clear, one-sentence description?
  • Failed tool calls return actionable error messages (not stack traces)?
  • Tool results are concise (no dumping raw HTML/JSON)?

3. Error Recovery

  • Retry logic with exponential backoff?
  • Graceful degradation (if tool X fails, try alternative)?
  • Never silent swallowing of errors?

4. Model Routing

  • Parsing/extraction tasks → cheapest tier (Haiku)?
  • Reasoning/planning → middle tier (Sonnet)?
  • Architecture/complex decisions → top tier (Opus)?
  • Never sending 100K tokens to the expensive model?

5. Memory Management

  • Important facts written to disk BEFORE compaction?
  • Memory files have clear structure (not just append-only)?
  • Cross-session memory actually loads at session start?

6. Planning

  • Uses TodoWrite or equivalent for multi-step coherence?
  • Each step has a testable definition of done?
  • Can explain what it's doing and why at any point?

Reference

Read docs/vision/VISION.md — search for "harness engineering evidence"

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.