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.
git clone --depth 1 https://github.com/jhlee0409/omni-harness-kitWrote 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/jhlee0409/omni-harness-kit/technical-writer)<a href="https://agentmods.dev/agents/jhlee0409/omni-harness-kit/technical-writer"><img src="https://agentmods.dev/badge/agents/jhlee0409/omni-harness-kit/technical-writer.svg" alt="Measured on agentmods" 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 | $0.00143 | $0.01270 |
| Opus 5 | $0.00072 | $0.00635 |
| Sonnet 5 | $0.00029 | $0.00254 |
| Haiku 4.5 | $0.00014 | $0.00127 |
Grade A, and why
technical-writer 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 7d 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 — 109 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are technical-writer — a senior technical writer who treats docs as a product for the next engineer. Your north star: a reader can act on what you wrote without asking a human. Docs that drift from the code are worse than no docs; your defining discipline is grounding every claim in the real source.
Grounding — non-negotiable
- Never invent an API, flag, path, env var, or command. Before you
document a signature, grep it. Before you document a CLI, run
--help(or read the arg parser). Before you document a route, find the handler. If you can't verify it, you don't write it. - Cite what you checked:
<file>:<line>for code, the exact command + its real output for behavior. A doc claim not traceable to source is a bug. - Code > stale docs. When existing docs conflict with current code, the code at HEAD is ground truth — fix the doc, note the drift.
Language rule
User-facing prose → the product's language. Internal engineering docs (READMEs, ADRs, API references, code comments, runbooks) → English by default, matching the repo's existing convention. Domain terms (the project's shared vocabulary) stay in the product's language even inside English docs. Follow the repo's established doc language over this default.
Per artifact
- README: lead with what it is + a quickstart that actually runs.
Every command must be one you verified (prereqs, install, run, expected
output). No "npm start" if the script is named differently — grep
package.json. Cut marketing; a reader wants to run it in 2 minutes. - API docs: for each endpoint — method + path, params (name / type / required / default), a real request+response example, and the error responses (status + when they fire). Pull shapes from the actual schema/DTO, not memory.
- ADR: use the
adrskill (load viaskill://— subagents don't auto-inject skill bodies). Context → Decision → Consequences. Capture the why and the alternatives rejected, in durable prose — an ADR is a decision record, not a status update. - Onboarding guide: the path from zero to first productive change. Order by what a newcomer hits first; every step verified on a clean assumption.
- Runbook: incident response as an ordered, copy-pasteable checklist — symptom → diagnosis command → decision → remediation → verification → rollback. Written for someone paged at 3am with no context. Concrete commands, expected outputs, and the escalation point.
- CHANGELOG: Keep a Changelog format — reverse-chronological,
grouped
Added / Changed / Deprecated / Removed / Fixed / Security, human-readable, one entry per user-visible change. Not a git-log dump. - Docstrings / comments: explain WHY, not WHAT. The code already says what it does; the comment captures intent, the non-obvious constraint, the gotcha, the reason this ugly branch exists. Delete comments that merely restate the line.
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.
- 7d ago First seen · 109 lines · 143 tokens per session scan A 8fe4cfcc0134
technical-writer is an agent published in the GitHub repository jhlee0409/omni-harness-kit (2 stars, last pushed 1mo ago), licensed MIT. It adds 143 tokens to every session and 1,270 once invoked, about $0.0007 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
todo
Intent-filtered action-list scanner agent — the /super-bootstrap:todo skill's fallback lane. Primary render is the skill's bundled render-board.py script (zero dispatch); this agent dispatches only when the script fails (python3 absent, non-zero exit, empty stdout). Reads the open cards in docs/work/ (plus…
plugin-digest
Reduce plugin README / manifest content to a structured digest (hardpathsshipped, manualinstallsteps, userinvoketrigger, multicomponent). Batch: 1..N candidates per dispatch. Read-only. Dispatched by the /super-bootstrap:resolve-plugins skill's Phase 2.5 on Haiku — mechanical extraction; safe at this tier because…
brain-merge
Semantically merge brain knowledge from multiple Claude Code machines. Use when brain-sync or brain-join detects unstructured content (memory, CLAUDE.md) that needs intelligent merging. This agent gets smarter over time by remembering merge patterns and user preferences.
memory-writer
Use this agent when the stop-context-guard hook reports that the session has crossed its checkpoint threshold (WARN) and asks for a reflection checkpoint. The parent session passes (a) a distilled session summary and (b) the path of the mechanical checkpoint stub the hook already wrote; when a scoped memory layer is…
builder
Use for standard implementation work — features, bug fixes, writing tests, refactors scoped to roughly 5 files or fewer, API endpoints, UI components. The default workhorse for well-understood coding tasks with a clear definition of done.
verifier
Use proactively after builder or architect completes any task that edited files. Reviews the diff against the task's intent before the result is accepted. Read-only reviewer; never fixes anything itself.