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.
npx agentmods add commands/whieet/harness-kit/evaluategit clone --depth 1 https://github.com/whieet/harness-kitWhat 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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00037 | $0.00415 |
| Opus 5 | $0.00018 | $0.00208 |
| Sonnet 5 | $0.00007 | $0.00083 |
| Haiku 4.5 | $0.00004 | $0.00042 |
Grade A, and why
evaluate 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 yesterday.
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.
What it actually says
/harness-kit:evaluate
Run an INDEPENDENT evaluation of the current change — the robust path for Generator/Evaluator separation (a generator grading itself is unreliable).
Steps
-
Read
.harness/config.jsonfirst and follow itslanguagepreference when relaying the evaluation to the user. Do not translate file/directory names, command names, or config keys. -
Dispatch the
evaluatorsubagent (via the Task/Agent tool, subagent typeevaluator). It runs with fresh context and has no edit tools, so it cannot grade its own work. Tell it what changed (the current diff scope / the active plan). -
The evaluator will: read
.harness/rubric.md+config.verificationRecipe, run each dimension's verification check (the project's MCP tools / test runners), score every dimension 1–5 (any dimension < 3 = FAIL), append its scores to.harness/state/trace.jsonl(feeds the config suggester), and return aVERDICT: PASS | WARN | FAIL. -
Relay the verdict. On
FAIL, list the must-fix items and do not declare the task done — revise and re-evaluate. OnWARN, pass but note the technical debt.
Notes
- This is the recommended, robust evaluator path. An optional experimental auto-fire variant (a
type:agentStop hook) exists inhooks/optional-auto-eval.json— it is context-only (cannot block), runs every Stop, and is off by default; see the README before enabling it. - The Stop-hook completion gate (
harness-verify+ plan DoD) remains the hard, blocking gate; the evaluator is the subjective-quality layer on top.
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.
- yesterday First seen · 23 lines · 37 tokens per session scan A e0bab3be9bcc
evaluate is a command published in the GitHub repository whieet/harness-kit (5 stars, last pushed 1mo ago), licensed MIT. It adds 37 tokens to every session and 415 once invoked, about $0.0002 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.
Other commands, from other repositories
verify
Adversarial spec-vs-implementation verification for a completed task. Dispatches the spec-mentor subagent with fresh context (no anchoring bias), parses its verdict (PASS / DRIFT / NEEDS-MARTY), and updates the verification queue. The v7.4.0 architectural replacement for a dedicated "mentor session.".
research
Enter RESEARCH mode for information gathering.
criar-skill
Use when creating new skills, automations, or specialized knowledge packages. Keywords: criar skill, nova skill, automatizar, conhecimento, TDD skill.
research
Delegate a thorough research investigation to the agy:runner subagent.
station
You are helping the user work with Station - the self-hosted AI agent orchestration platform.
delegate
Delegate investigation, an explicit fix request, or follow-up work to the Grok delegate subagent.