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/brekkylab/backlot/agents-mdgit clone --depth 1 https://github.com/brekkylab/backlotWhat 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 | $0.00759 | $0.00759 |
| Opus 5 | $0.00380 | $0.00380 |
| Sonnet 5 | $0.00152 | $0.00152 |
| Haiku 4.5 | $0.00076 | $0.00076 |
Grade A, and why
backlot 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 — 63 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md
Backlot serves enterprise SaaS APIs (Slack, Gmail, Google Drive, GitHub, Jira, and more) over a corpus the user supplies, with per-document ACLs. Fidelity to the real APIs is the point of the project — read this before changing anything.
Fidelity is measured, never assumed
- A divergence from the real vendor API is a bug. A claim about what the real API does needs a measurement against the real service or a quote from the vendor's spec — never memory, never another mock, never this repo's own earlier prose.
- When a validation rule disagrees with a value the code produces, do not assume the rule is wrong. Find which side has an external source first. Widening a pattern to make a test pass has shipped a real bug here before.
- Docstrings that attribute a shape to the real vendor are source attributions. Do not delete or invert them without checking the vendor's spec.
Where a change goes
backlot/routers/— one module per vendor: response shapes, status codes, pagination.backlot/schemas/*.schema.json— the record schema per source.docs/supported-sources.mdand the README source table are downstream of these.backlot/acl.py— which principal sees which document, in the vendor's own terms.backlot/importer/— how a corpus gets in: the bundled set, BYO JSONL,--id-map, the roster.tests/test_<source>.py— written against measured vendor responses.examples/— one self-contained script per service per integration.
Adding a source is never one file. The full checklist lives in issue #89 and is the same every
time: schema, router, ACL mapping, BYO field mapping and ids, tests, SDK example, regenerated
docs, pyproject.toml keywords.
Documentation rules (tests/test_docs.py enforces all of these)
- README.md stays ≤ 130 lines. If a change needs more room, the content belongs in
docs/. - Never state a source count in README.md. Counts go stale; the generated inventory carries the real one.
- Every relative link in every markdown file must resolve on disk. Do not link a path a stacked PR will add later.
docs/supported-sources.mdis generated. Never hand-edit between the generated markers — runpython scripts/gen_docs.py.- Do not describe the served surface as "read-only" in user-facing docs. The wording was removed deliberately.
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 · 63 lines · 759 tokens per session scan A b4fb4ade991a
backlot AGENTS.md is an instructions file published in the GitHub repository brekkylab/backlot (22 stars, last pushed 5d ago), licensed MIT. It adds 759 tokens to every session, about $0.0038 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
clevis AGENTS.md
AGENTS.md instructions for nazarli-shabnam/clevis, covering agents.md, architecture, database models (apps/api/src/core/db.py), job queue flow and rbac.
clevis copilot-instructions.md
Copilot instructions for nazarli-shabnam/clevis: See AGENTS.md — Copilot reads it natively.
clevis CLAUDE.md
Claude Code instructions for nazarli-shabnam/clevis, a project described as: Self-hosted GitHub security dashboard — tracks org security checks (MFA, branch protection, scanning) with score trends, plus automated Actions cache cleanup. FastAPI + Next.js + PostgreSQL.
clevis GEMINI.md
Gemini CLI instructions for nazarli-shabnam/clevis, a project described as: Self-hosted GitHub security dashboard — tracks org security checks (MFA, branch protection, scanning) with score trends, plus automated Actions cache cleanup. FastAPI + Next.js + PostgreSQL.
RepoGuardian-RHD-Enterprise AGENTS.md
AGENTS.md instructions for romil569/RepoGuardian-RHD-Enterprise, covering repoguardian agent notes, architecture, local commands, data backends and safety rules.
aegisai-enterprise-agent-platform AGENTS.md
AGENTS.md instructions for vpeetla-ai/aegisai-enterprise-agent-platform, covering agent instructions — vpeetla-ai org, agentic engineering (karpathy), stack awareness, repo conventions and skills repo.