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 deciqAI/knowledge-skills --skill cpa-return-reviewgit clone --depth 1 https://github.com/deciqAI/knowledge-skillsWrote 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/deciqai/knowledge-skills/cpa-return-review)<a href="https://agentmods.dev/skills/deciqai/knowledge-skills/cpa-return-review"><img src="https://agentmods.dev/badge/skills/deciqai/knowledge-skills/cpa-return-review/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/deciqai/knowledge-skills/cpa-return-review"><img src="https://agentmods.dev/badge/skills/deciqai/knowledge-skills/cpa-return-review.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 | $0.00188 | $0.02713 |
| Opus 5 | $0.00094 | $0.01357 |
| Sonnet 5 | $0.00038 | $0.00543 |
| Haiku 4.5 | $0.00019 | $0.00271 |
Grade A, and why
cpa-return-review 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 10d 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 — 170 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CPA — Pre-Delivery Return Review
Not legal or tax advice. No thresholds, limits, rates, or dates appear here — all of them expire. Verify every amount and every rule against current IRS/state guidance and current-year software logic. The practitioner who signs owns the return.
Activate when: a return is prepared and going to review; you are reviewing someone else's work; you want the same discipline applied by every reviewer in the firm. Do NOT activate when: you are preparing rather than reviewing; the question is one isolated treatment; the question is what an examiner would attack rather than whether the return is right — that is [tax-prep-pre-file-audit-premortem], on 1120-S as much as on 1040.
Why this skill
Errors that reach clients are rarely exotic. They are a carryover nobody re-checked, a state that quietly changed, a name spelled two ways, an election made last year and forgotten this year, and a number that is technically correct but obviously wrong next to last year's.
Reading a return top to bottom hunting for mistakes finds what you already suspect and nothing else. Different error classes hide from different kinds of attention, so the review is five separate passes with five different questions — not one careful read.
Process
Run in order. Each sweep is its own pass.
Sweep 1 — Identity and posture
Boring, and the source of the most embarrassing errors.
- Legal names match official records exactly, including suffixes and hyphenation
- Taxpayer identification numbers correct against the taxpayer's own records
- Tax year correct on every form and schedule
- Filing status reflects facts collected this year, not carried over
- Dependents current — ages, relationships and eligibility change annually
- Address current and consistent with the states being filed
- Entity name, identifier, formation state match the organizing documents (1120-S)
- Shareholder roster current, including mid-year changes (1120-S)
- Direct deposit / payment details verified this year, not carried forward
- Preparer identifiers and signature blocks complete
- Paid-preparer due-diligence checklist completed and retained for every covered credit and for head-of-household — see [tax-prep-8867-due-diligence]. A per-return penalty gate, and the one Sweep 1 item that costs money on its own rather than through a wrong number
- Taxpayer e-file authorization signed and dated for the correct return and year, before transmission
- Any consent required before disclosing or using taxpayer information obtained in the required form and sequence
- Every other required signature, election statement and attachment present
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.
- 10d ago First seen · 170 lines · 188 tokens per session scan A e3f2cf57b829
cpa-return-review is a skill published in the GitHub repository deciqAI/knowledge-skills (10 stars, last pushed 8d ago), licensed MIT. It adds 188 tokens to every session and 2,713 once invoked, about $0.0009 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 skills, from other repositories
invoice-chase
Runs collections on late invoices: builds an aging ladder (current/30/60/90+), assigns each invoice a rung on an escalation sequence with ready-to-send emails — friendly nudge, firm reminder with late-fee mention, final notice — plus a phone script, tone rules that preserve the relationship, a chase log, and…
plan-payroll
Plans payroll cash: true loaded cost per hire (gross plus employer taxes, benefits, tools), a payroll calendar with cutoffs and cash-out dates, a scenario table for new hire vs raise vs contractor-vs-employee, and a payroll-to-revenue check against rough industry bands. Use when the user asks "can I afford to hire"…
review-contract
Clause-by-clause contract review — inventories key clauses (term, termination, liability cap, indemnity, IP, payment, confidentiality, governing law, auto-renewal), flags deviations from market-standard positions with RED/YELLOW/GREEN severity and verbatim quotes, drafts redline language, and sets negotiation…
tax-prep
Organizes tax season — an organizer, not tax advice: a document checklist by category (income, expenses, assets, payroll, prior filings), an expense-category sweep that flags plausibly missed deductions in generic jurisdiction-agnostic categories, a quarterly-estimate calendar template, a clean handoff pack for the…
audit-support
Gets the books audit-ready: builds the PBC (prepared-by-client) list by area, drafts walkthrough narratives for key cycles (revenue, purchases, payroll), matches auditor samples to support, prepares tie-out schedules, tracks open items with owners and due dates, and drafts responses to findings. Use when the user says…
financial-statements
Builds a full statement set — P&L, balance sheet, and indirect cash flow — from a trial balance or transaction export, mapping accounts to a standard chart and proving every tie-out in code. Use when the user says 'build a P&L from this trial balance', 'turn this export into financials', 'I need a balance sheet for…