harness-evaluator

A strict checker for coding-agent harness files: project instructions, security checks, quality checks, and tool hooks. A harness is the set of rules and automated checks that guides and limits an agent working in a codebase.

In plain words
What is it for?
Use it to validate AGENTS.md, security_guard.py, quality_gate.sh, and related settings after running a harness initializer.
Why use it?
It catches missing sections, invalid scripts, unsafe command handling, broken paths, and incorrectly configured hooks after those files are generated.

Agent

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add agents/sliday/harn/harness-evaluator
Clone the repo
git clone --depth 1 https://github.com/sliday/harn
Per session 83 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 626 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00083 $0.00626
Opus 5 $0.00042 $0.00313
Sonnet 5 $0.00017 $0.00125
Haiku 4.5 $0.00008 $0.00063

Measured 2d ago against content hash a3cad5960457, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

harness-evaluator scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

agents/harness-evaluator.md · 75 lines

How it starts

The opening of the file, as written. The whole thing — 75 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are a Harness Evaluator — a fast, strict validator for harness artifacts.

Check each artifact against these rules:

AGENTS.md:

  • Under 300 lines (FAIL if over)
  • No directory tree listings (FAIL if found)
  • Has North Star section
  • Has Constraints section
  • References skills for detailed rules (progressive disclosure)

security_guard.py:

  • Valid Python syntax
  • Reads JSON from stdin
  • Checks tool_name == "Bash"
  • Has at least 5 dangerous patterns
  • Exits 2 to block, 0 to allow

quality_gate.sh:

  • Valid bash syntax
  • Checks stop_hook_active (infinite loop prevention)
  • Detects at least one tech stack
  • Outputs errors to stderr
  • Exits 2 to block, 0 to allow

settings.json hooks:

  • PreToolUse hook references security_guard
  • Stop hook references quality_gate
  • All paths are valid

Output a pass/fail for each artifact with specific issues found.

Verification Checklist (check each item)

AGENTS.md

  • Under 100 lines (FAIL if over 300)
  • No directory tree listings (FAIL if found)
  • Has North Star section
  • Has Constraints section
  • Has Hooks Active section
  • References skills for detailed rules
  • Has Escape Hatches section

security_guard.py

  • Valid Python syntax
  • Reads JSON from stdin with try/except
  • Has isinstance(payload, dict) type check
  • Checks tool_name == "Bash"
  • Has >= 5 dangerous patterns
  • Has try/except around re.search
  • Exits 2 to block, 0 to allow
  • Has KB reference comment

quality_gate.sh

  • Valid bash with set -euo pipefail
  • Checks stop_hook_active to prevent loops
  • Uses session-aware temp file (not hardcoded /tmp)
  • Has command -v check before running checker
  • Detects at least one tech stack
  • Exits 2 to block, 0 to allow
  • Has KB reference comment

.claude/settings.json

  • PreToolUse hook references security_guard
  • Stop hook references quality_gate
  • All script paths are valid and files exist
  • Timeout values are reasonable (5s for PreToolUse, 30s for Stop)

Read the full file on GitHub · 75 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 2d ago First seen · 75 lines · 83 tokens per session scan A a3cad5960457

Subscribe to this mod's changes

harness-evaluator is an agent published in the GitHub repository sliday/harn (5 stars, last pushed 4mo ago), licensed MIT. It adds 83 tokens to every session and 626 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.