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/kunchenguid/backpass/agents-mdgit clone --depth 1 https://github.com/kunchenguid/backpassWhat 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 | $0.05264 | $0.05264 |
| Opus 5 | $0.02632 | $0.02632 |
| Sonnet 5 | $0.01053 | $0.01053 |
| Haiku 4.5 | $0.00526 | $0.00526 |
Grade A, and why
backpass 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 2d 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 — 272 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Project agent memory
backpass is an npm CLI that runs a "backward pass" over a repo's agent memory files: it
discovers past agent-session transcripts tied to the repo, analyzes them, and proposes
evidence-backed edits to AGENTS.md / CLAUDE.md under a token budget.
Orientation
VISION.mdis the acceptance policy: what a change must be true to, and its closing accept/resist tests.VISION-ANSWERS.mdrecords the hypotheticals the author ruled on and why, so a change that argues one of those calls should go the other way starts there.README.mddocuments the user-facing surface;src/cli.jsis the authoritative flag list.- The pipeline is one stage per module, in order:
src/discovery/->src/sample.js(cap) ->src/distill.js->src/analyze.js->src/consolidate.js(gap-identity merge, insidefoldForRun) ->src/fold.js->src/synthesize.js(withsrc/workspace.js+src/diff.js) ->src/proposal.js->src/apply/. Each module's header comment explains its role; read those before changing a stage. - User-facing step names are the training-loop terms, not the module names: discover =
"collect samples", analyze = "calculate loss", fold = "aggregate gradients", synthesize =
"gradient descent" (
STAGE_LABELSinsrc/tui/render.js). Internal keys, event names, and the typed subcommands (scan,analyze,propose) keep their short names; only what the user reads changes. Never introduce a multi-word subcommand. - Zero runtime dependencies, ESM, no build step. Node >= 22.5 (for
node:sqlite). - pnpm is the package manager;
pnpm run checkruns lint, format:check, typecheck, and tests. All tests are offline and use fixtures undertest/fixtures/. Supply-chain settings (release-age cooldown, build-script deny) live inpnpm-workspace.yaml. - Releases are automated by release-please (
.github/workflows/release-please.yml, npm trusted publishing). Never hand-editCHANGELOG.mdor.release-please-manifest.json; CI guards reject PRs that touch them. .github/workflows/no-mistakes-required.ymlis a thin caller of the sharedkunchenguid/no-mistakes/.github/actions/require-no-mistakescomposite action, pinned to an immutable commit SHA and never@main. Enforcement logic and tests live upstream in the no-mistakes repository; change enforcement there, never by copying it locally, and bump the pin in a deliberate separate PR. This repo still owns theon:,concurrency,permissions, job name, and author-exemptionif:. The action binds the attestation to the head SHA, so a PR whose body no-mistakes did not rewrite for the current head goes red by contract: push throughgit push no-mistakes.
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.
- 2d ago First seen · 272 lines · 5,264 tokens per session scan A 1d39066700fa
backpass AGENTS.md is an instructions file published in the GitHub repository kunchenguid/backpass (600 stars, last pushed 2d ago), licensed MIT. It adds 5,264 tokens to every session, about $0.0263 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
codex AGENTS.md
AGENTS.md instructions for openai/codex, covering rust/codex-rs, the codex-core crate, code review rules, crate api surface and model visible context.
vscode buildNext.instructions.md
Working notes and architecture documentation for the new esbuild-based build system in build/next. Use when making changes to the new build pipeline (transpile/bundle commands, NLS plugin, source-map handling, resource copying, or self-hosting watch tasks).
next.js AGENTS.md
Instructions for vercel/next.js, covering next.js development guide, codebase structure, monorepo overview, core package: packages/next and other important packages.
vscode oss-third-party-notices.instructions.md
Instructions for microsoft/vscode, covering vs code oss third-party-notices pipeline, architecture, pipeline flow in ci, applying the notice (cutover) and fallback chain (never fail the build).
spec-kit AGENTS.md
AGENTS.md instructions for github/spec-kit, covering agents.md, about spec kit and specify, quickstart — add a new integration in 5 steps, integration architecture and integrationmanifest — file tracking.
langchain AGENTS.md
AGENTS.md instructions for langchain-ai/langchain, covering global development guidelines for the langchain monorepo, corridor security analysis, project architecture and context, monorepo structure and development tools & commands.