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 agentmods add instructions/djouallah/duckrun/agents-mdgit clone --depth 1 https://github.com/djouallah/duckrunWhat 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 | $0.00631 | $0.00631 |
| Opus 5 | $0.00316 | $0.00316 |
| Sonnet 5 | $0.00126 | $0.00126 |
| Haiku 4.5 | $0.00063 | $0.00063 |
Grade A, and why
duckrun AGENTS.md 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 3d 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 — 48 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md
Guidance for AI agents working in this repo or helping someone use duckrun.
duckrun is a dbt adapter: DuckDB executes the model SQL, delta-rs writes the result as a Delta Lake table, and dbt orchestrates the DAG. It runs the same on a laptop, in CI, or in a Fabric notebook.
Before helping with a duckrun project, read the skill — it is the authoritative guide:
plugins/duckrun-projects/skills/duckrun-projects/SKILL.md
A few defaults differ from other dbt adapters and cause silent mistakes if you assume the usual behavior:
threads:is honored (dbt's default is 1), but every model writes a real table: concurrent writers share one DuckDBmemory_limitand a microbatch model's batches always run in order. More threads help many network-bound models; they don't help one big merge.- DuckDB is in-memory; there is no database file. The Delta tables are the only state.
- Incremental strategy defaults depend on
unique_key(mergewith it,appendwithout). For large tables,mergevs a dedup-in-SQLappend(auto-fenced when the model reads{{ this }}) matters a lot. - OneLake/Fabric auth is just a bearer token; paths use lakehouse GUIDs, not names.
Consult the SKILL.md before writing profiles.yml or any incremental model.
Working on duckrun itself
Read CONTRIBUTING.md first — it has the full flow and the rules. The
short version:
- Every change lands via a pull request, including the owner's own. Branch off
main(fix/,feat/,docs/,ci/,test/), push the branch, open a PR. This repo overrides any general "you own this repo, so commit straight tomain" rule you may be carrying — there is no direct-push path here. - Never push to
main, and never touch thelegacybranch. Two things stay onmainand are not to be "fixed": CI's own[skip ci]scorecard commits, and the releasevX.Y.Ztag push. - Never modify a test to make a PR pass. No
skip, noxfail, no weakened assertion. A red test means fix the adapter. - Don't add new public API without discussing it first. Agents are the ones most likely to invent surface nobody asked for. A question is not a work order.
- It's fine to run tests locally for this project.
- Keep duckrun boring: prefer the dull, idiomatic solution. The only intentional bet is the Arrow + delta-rs core, because there's no alternative way to write Delta.
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.
- 3d ago First seen · 48 lines · 631 tokens per session scan A e2ea7fb0753c
duckrun AGENTS.md is an instructions file published in the GitHub repository djouallah/duckrun (48 stars, last pushed 5d ago), licensed MIT. It adds 631 tokens to every session, about $0.0032 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 2026-08-30.
Other instructions, from other repositories
pgContext AGENTS.md
Instructions for Evokoa/pgContext, covering agents.md — installing & using pgcontext with an ai agent, what pgcontext is (and why it's worth using), environment facts (pins — do not guess), path a — docker (preferred; zero build, most deterministic) and path b — build from source (when docker is unavailable).
flock copilot-instructions.md
Instructions for dais-polymtl/flock, covering copilot instructions for flock, repository layout, building, setup vcpkg (first time or after clean) and release build.
pdf-brain AGENTS.md
Instructions for joelhooks/pdf-brain, covering pdf-brain agent notes, libsql quirks, ai sdk pattern, key files and docs.
markdown-vdb CLAUDE.md
Instructions for geckse/markdown-vdb, covering markdown vdb, architecture, project structure, core design decisions and key conventions.
MinSync AGENTS.md
Instructions for NomaDamas/MinSync, covering minsync and core principles.
astra-knowledge-base-mcp AGENTS.md
AGENTS.md instructions for alrcatraz/astra-knowledge-base-mcp, covering astra-knowledge-base-mcp — agent guide, project overview, code map, llm wiki interop and development principles.