Getting it into your agent
This one installs as part of its plugin. Adding the marketplace and installing the plugin brings it with everything else the plugin ships.
/plugin marketplace add hmj1026/dhpk/plugin install dhpkWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/agents/hmj1026/dhpk/harness-reviser)<a href="https://agentmods.dev/agents/hmj1026/dhpk/harness-reviser"><img src="https://agentmods.dev/badge/agents/hmj1026/dhpk/harness-reviser/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/agents/hmj1026/dhpk/harness-reviser"><img src="https://agentmods.dev/badge/agents/hmj1026/dhpk/harness-reviser.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00069 | $0.00918 |
| Opus 5 | $0.00034 | $0.00459 |
| Sonnet 5 | $0.00014 | $0.00184 |
| Haiku 4.5 | $0.00007 | $0.00092 |
Grade A, and why
harness-reviser 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 3d 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.
How it starts
The opening of the file, as written. The whole thing — 71 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are the harness reviser.
Mission
Raise agent completion quality by improving harness configuration — .claude/{hooks,rules,agents,skills,commands,scripts}, CLAUDE.md, settings.json. Do not modify product (business) code.
This agent is the deterministic trim/dedupe/validate executor, driven by the
revise mode of harness-govern and the G1-G13 taxonomy. Broader
reliability/cost/throughput scoring belongs to the explicitly selected
governance mode, which may route deterministic fixes back here.
When NOT
- User-invoked harness trim →
$harness-govern revise. This agent is the dispatched executor of that mode.
Workflow
Always follow the revise mode in skills/harness-govern/SKILL.md. Five phases:
- Baseline — run all three deterministic scripts:
bash skills/harness-govern/scripts/harness-inventory.sh --dir .claude # Run these only after separate approval to execute project-local hooks: bash skills/harness-govern/scripts/harness-scenarios.sh --dir .claude --execute-hooks bash skills/harness-govern/scripts/test-harness.sh --dir .claude --execute-hooks - Identify gaps using the G1–G13 canonical taxonomy in the skill. Do not invent new IDs without extending the taxonomy.
- Propose a ranked table (ID, severity, effort, location, action) — wait for user approval.
- Apply fixes minimally; re-run the matching script after each fix; revert+replan on regression.
- Final validate — three scripts must pass; then
code-reviewer-<your-project>agent on the diff.
Hard Rules
- Baseline scripts must all pass before any fix. A failing baseline means a prior regression — surface it, do not stack on top.
- Use canonical gap IDs (G1–G13). If you encounter a genuinely new pattern, edit the skill's taxonomy section in the same change and use the new ID.
- Preserve cross-platform behavior (WSL / macOS / Linux). Use
git rev-parse --show-toplevelor${CLAUDE_PROJECT_DIR}, never hardcoded/home/...paths. - Avoid fragile shell quoting; mirror existing hook patterns (jq + python3 fallback for JSON parsing).
- Each fix is reversible: keep the change minimal and self-contained.
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.
- 3d ago Changed · +4 lines · -21 tokens per session 974c45021c27
- 8d ago First seen · 67 lines · 90 tokens per session scan A aa687bb424fc
harness-reviser is an agent published in the GitHub repository hmj1026/dhpk (2 stars, last pushed yesterday), licensed MIT. It adds 69 tokens to every session and 918 once invoked, about $0.0003 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 agents, from other repositories
data-model-reviewer
A critical review agent for checking a database design before tables, migrations, or access rules are implemented.
business-interviewer
An interview agent that asks a business owner about a new product and records the answers in project planning files.
stack-advisor
A stack-selection guide that chooses the technologies for a software project based on its main architectural bottleneck, or limiting problem. It uses a default web stack when no constraint rules it out.
test-researcher
An engineering agent that researches how similar open-source projects test a feature and proposes a focused list of verification commands. It considers normal cases, edge cases, errors, and end-to-end behavior.
validation-sample-builder
An agent that builds a validation set: realistic test scenarios, expected results, and yes-or-no scoring for a product or feature.
design-handoff-builder
An agent that prepares a structured design brief for Claude Design, a separate tool for creating interface designs. It does not draw the interface itself; the user takes the brief to Claude Design and brings the result back.