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/coltonbearden/carrelWrote 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/commands/coltonbearden/carrel/refs)<a href="https://agentmods.dev/commands/coltonbearden/carrel/refs"><img src="https://agentmods.dev/badge/commands/coltonbearden/carrel/refs/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/commands/coltonbearden/carrel/refs"><img src="https://agentmods.dev/badge/commands/coltonbearden/carrel/refs.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 | $0.00030 | $0.00966 |
| Opus 5 | $0.00015 | $0.00483 |
| Sonnet 5 | $0.00006 | $0.00193 |
| Haiku 4.5 | $0.00003 | $0.00097 |
Grade A, and why
refs 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 today.
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.
What it actually says
Find the reference numbers the user asked about: $ARGUMENTS
Run the carrel CLI via Bash. Map the request onto the real flags in the --help block below (regenerated from the CLI by scripts/sync_plugins.py; if the installed carrel refs --help differs, trust the installed version — never invent flags):
Usage: carrel refs [OPTIONS] PATHS...
Find reference numbers (invoice, PO, IBAN, routing, tracking, …) in PATH...
Directories are walked like `index` (hidden and .gitignored entries skipped; images only with
--ocr). Every supported file type works; the text comes from the same spine `pack` and `index`
use. Values with a check digit (iban, routing, isbn, gtin, cc) are reported only when it verifies.
JSON output is a list of {path, refs: [{kind, value, count, valid, pages, lines}]} — or, with
--link, [{kind, value, files, count}]. Kinds are shared with `redact --builtin`.
Options:
--kind K1,K2 Only these kinds, comma-separated. Default: every reference and identifier
kind (invoice, po, order, check, account, tracking, ticket, iban, routing,
ein, vat, isbn, gtin, doi, ups, usps); PII kinds (email, phone, ssn, ipv4,
cc) only when named.
--pattern NAME=REGEX Extra kind to look for (repeatable). A (?P<v1>…) group is the value;
otherwise the whole match is.
--tag Tag each file in the desk db under --root with ref:<kind>:<value>.
--link Group by reference instead of by file: which files share each value.
--all With --link, also list values seen once.
--ocr OCR images and scanned PDFs (needs tesseract / ocrmypdf).
--fail-empty Exit 5 when no reference was found.
--json Machine-readable JSON output.
--help Show this message and exit.
- Start with
carrel --json refs PATH...on the files or folder in question; every supported type works (PDF, docx, xlsx, md, txt, eml, …) and PDFs report page numbers. - "Which files mention invoice X" / "what belongs together" →
--link(values seen in more than one file;--alllists every value). - "Tag them" / "link them in the desk" →
--tagwritesref:<kind>:<value>tags into the desk db under the global--root(default cwd); afterwardscarrel tag find ref:invoice:inv-2026-0042andcarrel search QUERY --tag ref:...find every document carrying that reference. - Narrow with
--kind invoice,po,iban; the PII kinds (email, phone, ssn, ipv4, cc) are only scanned when named. Add a house format with--pattern NAME=REGEX(a(?P<v1>…)group is the value). - Values with a check digit (IBAN, routing, ISBN, GTIN, card) are only reported when it verifies (
"valid": true); other kinds carry"valid": null. - Scanned PDFs and images need
--ocr(tesseract); a missing binary exits 3 with the install hint — report it rather than guessing.
Report what was found conversationally: per file, the kinds and values (with pages), then any shared references. The same kinds are available to /carrel-documents:redact --builtin when the user wants them removed instead.
Requires the carrel CLI on PATH. If carrel is not found, tell the user to install it with uv tool install carrel (see the repo's INSTALL notes), or run it as uv run carrel ... from the carrel repo root.
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.
- today First seen · 51 lines · 30 tokens per session scan A d39915cf6ddb
refs is a command published in the GitHub repository coltonbearden/carrel (1 stars, last pushed today), licensed MIT. It adds 30 tokens to every session and 966 once invoked, about $0.0002 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 2026-09-11.
Other commands, from other repositories
clean
Turn messy or multi-format CSV exports into one clean, categorized, analysis-ready file.
payslip
Upload and process Danish payslips (lønsedler) for pension tracking, income analysis, and enhanced financial advice.
receipt
Upload and process receipt images or PDF invoices, extract line items, and match to bank transactions.
sheet-report
Analyze a Numbers or Excel file and generate a report with charts.
numbers-to-excel
Convert a Numbers document to Excel.
add-skill
Add a new Skill to the current plugin with proper structure.