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/yuri-semenenko/ai-engineering-workspace/securitygit clone --depth 1 https://github.com/yuri-semenenko/ai-engineering-workspaceWhat 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 | $0.00751 | $0.00751 |
| Opus 5 | $0.00376 | $0.00376 |
| Sonnet 5 | $0.00150 | $0.00150 |
| Haiku 4.5 | $0.00075 | $0.00075 |
Grade A, and why
Security Standards 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 — 50 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Security Standards
Apply when a change touches authentication, authorization, user input, external calls, secrets, payments, personal data, file uploads, database access policies, or server-side actions.
Rationalizations
Excuses that precede a skipped control, paired with the answer. If you think the left, the right applies.
- "Internal endpoint, no auth needed." -> Internal is a network assumption, not a guarantee. Authn + authz still apply.
- "Just an MVP, harden later." -> Auth, secrets, and injection are table stakes, not later-work.
- "Input comes from our own frontend." -> The frontend is not a trust boundary. The API is. Validate at the boundary.
- "It's behind a login, so it's safe." -> Authn is not authz. IDOR lives here: check owner/role on every resource.
- "The framework auto-escapes output." -> Verify the sink. Raw HTML, raw SQL, and template bypasses exist.
Check
- Trust boundaries: know the source of every input and whether it crosses from user, browser, webhook, third party, or internal service into privileged code.
- Authorization on the server, not just gated UI. Verify actor, tenant, ownership, role, and object-level access before reading or mutating data.
- Input validation at the boundary: shape, type, range, encoding, and allowed values before use in SQL, file paths, shell, redirects, templates, or API calls. Prefer allowlists.
- Secrets stay out of code, logs, and responses. Never print tokens, connection strings, cookies, private keys, or signed URLs.
- Output safety: do not leak internal errors, stack traces, access tokens, personal data, or other tenants' data in responses, logs, or analytics.
- Postgres/Supabase: row-level security on every table, least-privileged client on the server (anon vs service-role), and ownership checks not left to RLS alone in trusted server paths.
- Dependencies: prefer existing ones. For new packages check maintenance, transitive risk, and postinstall scripts, and watch for typosquatting.
- LLM/AI calls: treat model output as untrusted (no direct eval, SQL, shell, innerHTML, or file paths), enforce permissions in code rather than the system prompt, and keep secrets and cross-user data out of the context window.
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 · 50 lines · 751 tokens per session scan A cbc19dfb0296
Security Standards is an instructions file published in the GitHub repository yuri-semenenko/ai-engineering-workspace (1 stars, last pushed 4d ago), licensed MIT. It adds 751 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 2026-08-31.
Other instructions, from other repositories
apm architecture.instructions.md
Single canonical owner discipline: one authority per durable decision, guarded by a regression test + a static boundary check.
spec-kitty AGENTS.md
AGENTS.md instructions for Priivacy-ai/spec-kitty, covering spec kitty development guidelines, ⚠️ critical: load the project charter first, ⚠️ critical: template source location, ⚠️ critical: use canonical sources, never improvise and ⚠️ critical: git workflow — no direct pushes to origin/main.
apm python.instructions.md
Python development guidelines.
mcp-server-excel excel-com-interop.instructions.md
Instructions for sbroenne/mcp-server-excel, covering excel com interop, api selection, batch execution and exceptions, com lifetime and excel application state.
mcp-server-excel coverage-prevention-strategy.instructions.md
Instructions for sbroenne/mcp-server-excel, covering generated surface coverage, contract change workflow, required checks and common incomplete changes.
mcp-server-excel mcp-llm-guidance.instructions.md
Instructions for sbroenne/mcp-server-excel, a project described as: Automate real Microsoft Excel with AI via MCP Server or CLI — Power Query, DAX, VBA, PivotTables, charts, and 326 operations.