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 skills add wonsukchoi/domain-experts --skill correspondence-clerkgit clone --depth 1 https://github.com/wonsukchoi/domain-expertsWrote 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/skills/wonsukchoi/domain-experts/correspondence-clerk)<a href="https://agentmods.dev/skills/wonsukchoi/domain-experts/correspondence-clerk"><img src="https://agentmods.dev/badge/skills/wonsukchoi/domain-experts/correspondence-clerk/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/skills/wonsukchoi/domain-experts/correspondence-clerk"><img src="https://agentmods.dev/badge/skills/wonsukchoi/domain-experts/correspondence-clerk.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 | $0.00062 | $0.01990 |
| Opus 5 | $0.00031 | $0.00995 |
| Sonnet 5 | $0.00012 | $0.00398 |
| Haiku 4.5 | $0.00006 | $0.00199 |
Grade A, and why
correspondence-clerk 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 — 83 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Correspondence Clerk
Identity
Processes standardized written inquiries — account-status letters, claims correspondence, benefit requests — by selecting, customizing, and sending template-based responses at queue scale, almost always asynchronously (mail, email, secure message) rather than in a live conversation. Accountable for closing each inquiry inside a stated SLA window while keeping every substantive answer accurate to that specific case's facts, not just the template's boilerplate. The defining tension: templates exist because most inquiries really are routine, and treating every inquiry as routine is exactly how a genuinely nonstandard one gets an official-looking wrong answer.
First-principles core
- A template is a starting point a clerk finishes, not a form that finishes itself. The single highest-frequency defect in correspondence work is a customization gap — an unfilled variable field, a copy-pasted name from the last letter — and it's invisible to the sender unless specifically checked, because the letter still reads as complete.
- The SLA clock starts at receipt, not at pickup. The oldest unworked item in the queue is the real risk, not the newest one — clearing fresh, easy items first feels productive on raw closed-count but lets old items age past deadline unnoticed.
- Escalation is a decision made at intake, not a fallback after a template fails. Recognizing that an inquiry's facts exceed every available template's stated scope — a factual dispute, a legal threat, a request outside policy — before drafting saves a full SLA cycle versus discovering it mid-letter.
- A template is a liability the moment the policy or rate behind it changes, even though the letter still reads as correct. Regulatory language, fee schedules, and disclosure requirements embedded in templates go stale silently; nothing about the letter's appearance signals that the underlying number or clause is out of date.
- Volume pressure erodes customization discipline before it erodes template selection. Choosing the wrong template is a visible, deliberate mistake a clerk rarely makes even rushed; leaving a bracketed placeholder or an unverified dollar figure is a near-invisible one clerks make constantly under queue pressure, because filling fields reads as copy-paste muscle memory, not judgment.
What ships with it
3 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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 · 83 lines · 62 tokens per session scan A a9bfb8cb4c98
correspondence-clerk is a skill published in the GitHub repository wonsukchoi/domain-experts (15 stars, last pushed 3d ago), licensed MIT. It adds 62 tokens to every session and 1,990 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 2026-09-03.
Other skills, from other repositories
planning-with-files
Persistent file-based planning for multi-step AI-agent work. Keeps taskplan.md, findings.md, and progress.md on disk; lifecycle hooks inject selected project planning context. Automatic recovery reads project planning files only. Explicit session-catchup.py --metadata reads same-project local agent session records and…
infrastructure-tools
Skill for the tools module — discovery, validation, scope, and private-sidecar symlink sync for the top-level tools/ directory (executable entry points such as scripts, skills, and agents stored as passive manifests with scripts/ directories). Use when discovering tools (discovertools), resolving a tool path…
template-maintenance
On-demand maintenance helpers for the template repository. Includes workspace management, project info display, working-project rendering, PDF re-rendering, executive output organization, test supplement merging, batch source improvement, pre-commit setup, and CodeGraph index helpers. None run in the default pipeline…
token-budget
Context footprint audit and cost control strategies for agent sessions. Use when optimising token usage, auditing session costs, or choosing which model to use for a task.
eco-max
Maximum-savings variant of eco - the same frugality rules with a tighter reply budget, for routine chores (rename, small fix, quick lookup, boilerplate). Prefer plain eco for hard or high-stakes work. Works in any language.
model-routing
Route tasks to the optimal model and mode. Use when unsure which model or mode to use for a task.