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/teimurjan/blazediff/agents-mdgit clone --depth 1 https://github.com/teimurjan/blazediffWrote 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/instructions/teimurjan/blazediff/agents-md)<a href="https://agentmods.dev/instructions/teimurjan/blazediff/agents-md"><img src="https://agentmods.dev/badge/instructions/teimurjan/blazediff/agents-md.svg" alt="Measured on agentmods" 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 | $0.00910 | $0.00910 |
| Opus 5 | $0.00455 | $0.00455 |
| Sonnet 5 | $0.00182 | $0.00182 |
| Haiku 4.5 | $0.00091 | $0.00091 |
Grade A, and why
blazediff 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 4d 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 — 39 lines — stays where its author put it; the contents beside it link to each section on GitHub.
BlazeDiff Agent Rules
Monorepo for image and object diffing. pnpm workspaces. Rust core in crates/blazediff/, TS packages in packages/, agent in packages/agent/.
Hard rules
- No em dashes. Anywhere. Commits, PRs, code comments, docs, chat output. Use a period, comma, parens, or colon.
- Never edit
.blazediff/manifest.jsonby hand. Drive it through the agent CLI. - Never skip pre-commit hooks (
--no-verify,--no-gpg-sign). Fix the underlying issue. - Never force-push to
mainor amend published commits. - Keep these three in sync; edit all of them together:
skill/blazediff/SKILL.md(authoritative playbook)apps/website/app/docs/agent/page.mdxpackages/agent/README.md
- Don't add watershed or distance-transform to the interpret pipeline. Morph close + connected components is sufficient.
- Don't restructure interpret into atomic-regions / semantic-groups / score-labels. Improve evidence extraction, not topology.
- Don't add
deno.jsonto@blazediff/core-native-*,@blazediff/bun,@blazediff/vitest,@blazediff/jest,@blazediff/ui,@blazediff/react. They stay NPM-only. - Verify JSR slow-types per package (
cd packages/<pkg> && npx jsr publish --dry-run). Never from the workspace root. - When iterating on diff algorithms, use
--fixtures=<small-subset>and--iterations=2. Run the full suite only before merging. - After Rust changes, verify both binding feature builds for the crate you touched:
cargo check -p <crate> --features napiand--features python(blazediff, blazediff-ssim and blazediff-interpret each ship both). - JSR-only Node imports (e.g.
node:bufferin@blazediff/core) live inpackages/<pkg>/jsr.patch. Don't commit source with the patch applied;scripts/checks/check-jsr-patches-clean.shenforces this in pre-commit.
Quick commands
pnpm build(excludes website),pnpm test,pnpm typecheck,pnpm deno:testpnpm check:write(biome lint + format)npx @j178/prek run --all-files(full pre-commit suite)cd crates && cargo test -p blazediff(Rust core)
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.
- 4d ago First seen · 39 lines · 910 tokens per session scan A 9e1b5d29c8fe
blazediff AGENTS.md is an instructions file published in the GitHub repository teimurjan/blazediff (337 stars, last pushed 5d ago), licensed MIT. It adds 910 tokens to every session, about $0.0046 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
json-difference CLAUDE.md
Instructions for lukascivil/json-difference, covering claude.md, general guidelines for working with nx, scaffolding & generators, when to use nxdocs and repository overview.
json-difference AGENTS.md
Instructions for lukascivil/json-difference, covering general guidelines for working with nx, scaffolding & generators and when to use nxdocs.
GitWand AGENTS.md
AGENTS.md instructions for devlint/GitWand, covering agents.md — gitwand ai agent rules, monorepo structure, security — critical rules, no shell string interpolation in git commands and never bypass saferepopath().
critique AGENTS.md
AGENTS.md instructions for remorses/critique, covering opentui, using unreleased opentui versions, bun, react and rules.
rtl-agent-team CLAUDE.md
Claude Code instructions for babyworm/rtl-agent-team, covering language rule, important — project identity, why agentic coding for silicon ip, plugin runtime vs. development context (critical) and plugin architecture: dynamic prompt injection.
r2-web CLAUDE.md
Instructions for vikiboss/r2-web, covering claude.md, 项目速览, 或, 代码地图 and 快速定位表.