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 michielinksee/bantou --skill nightly-batchgit clone --depth 1 https://github.com/michielinksee/bantouWrote 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/michielinksee/bantou/nightly-batch)<a href="https://agentmods.dev/skills/michielinksee/bantou/nightly-batch"><img src="https://agentmods.dev/badge/skills/michielinksee/bantou/nightly-batch/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/michielinksee/bantou/nightly-batch"><img src="https://agentmods.dev/badge/skills/michielinksee/bantou/nightly-batch.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 | $0.00031 | $0.00663 |
| Opus 5.5 | $0.00012 | $0.00265 |
| Sonnet 5.5 | $0.00006 | $0.00133 |
| Haiku 4.5 | $0.00003 | $0.00066 |
Grade A, and why
nightly-batch 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 2mo 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 — 97 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Nightly Batch
Batch-process all client companies overnight. Fetches unprocessed transactions from freee API, runs the full classification pipeline, and generates a summary for morning review.
When to use
Invoke this skill when:
- The firm wants to automate daily transaction processing
- Setting up scheduled overnight runs via Anthropic Routines
- Manually triggering a batch run for all or selected clients
Pipeline
The nightly_run tool executes the following steps for each client:
1. Fetch unprocessed transactions
- Calls
list_freee_companiesto enumerate all connected companies - For each company, calls
list_freee_dealswith a date filter for unprocessed entries since the last run
2. Exclusion filter
Each transaction passes through Stage 0 exclusion (7 rules). Excluded items are logged but not classified.
3. Memory recall
Query the correction memory for each transaction description. Memory hits are classified instantly without API cost.
4. Keyword + AI classification
Remaining items flow through Stage 1 keyword matching, then Stage 2 AI fallback for unmatched entries.
5. Confidence routing
Results are routed by confidence level:
| Confidence | Action | Review needed |
|---|---|---|
| High (>= 85%) | Auto-classified, logged | No |
| Medium (70-84%) | Auto-classified, flagged | Morning review |
| Low (< 70%) | Held for human review | Yes |
6. Summary generation
After all clients are processed, a batch summary is produced:
- Total clients processed
- Total transactions classified
- Breakdown by confidence tier
- Items requiring morning review (with details)
- API token usage for the run
Scheduling with Anthropic Routines
The nightly batch is designed to work with Anthropic Routines for automated scheduling. Typical configuration:
- Schedule: 02:00 JST daily (weekdays only)
- Trigger: Cron expression or Routines scheduler
- Notification: Summary posted to the firm's Slack or email
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.
- 2mo ago First seen · 97 lines · 31 tokens per session scan A 19d35abdfdea
nightly-batch is a skill published in the GitHub repository michielinksee/bantou (2 stars, last pushed 2mo ago), licensed MIT. It adds 31 tokens to every session and 663 once invoked, about $0.0001 per session on Opus 5.5. 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
accounting-audit-system-builder
Routes an accounting or audit request to the right module skill, from software selection through monthly closing, asking only what is missing. Use for books of accounts or audit files.
bookkeeping
Use when a small business needs audit-ready books — a chart of accounts, posting a transaction to the right account and side, clearing an uncategorized bank feed, cash vs accrual, or a ledger that won't tie to the bank. NOT interpreting the numbers — runway, burn, P&L cadence (that is finance-ops), NOT issuing…
countbean-accounting
Rules for writing correct double-entry Beancount transactions when managing a Countbean cloud book — account naming, how to balance postings, categorisation, and safe ingestion. Use whenever recording, importing, or correcting financial data in a Countbean book.
jaz-cli
Use this skill when running Clio CLI commands, building shell scripts with Clio, debugging auth issues, understanding --json output, paginating results, or chaining multi-step accounting workflows from the terminal. Covers auth precedence, output formats, entity resolution, and common workflow patterns. Also use when…
jaz-recipes
Use this skill when modeling complex multi-step accounting transactions: anything that spans multiple periods, involves changing amounts, or requires linked entries. Covers 16 IFRS-compliant recipes (prepaid amortization, deferred revenue, loans, IFRS 16 leases, hire purchase, fixed deposits, asset disposal, FX…
jangbu-jongso
A Korean tax-document checklist generator for individuals, freelancers, sole-company owners, mixed earners, and certain business owners.