Synapsor-Runner AGENTS.md

Repository instructions for Synapsor Runner, a safety layer that lets database agents use reviewed business actions instead of direct SQL against PostgreSQL or MySQL.

In plain words
What is it for?
Use it when adding or changing database capabilities, approvals, writeback behavior, tenant controls, or safety tests.
Why use it?
It protects databases by requiring proposals and outside approval before writes, while checking scope, conflicts, affected rows, and replay evidence.

Instructions file for CodexOpenCode

Install

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.

agentmods
npx agentmods add instructions/synapsor/synapsor-runner/agents-md
Clone the repo
git clone --depth 1 https://github.com/Synapsor/Synapsor-Runner

Made for: Codex, OpenCode.

Per session 972 This file is loaded in full into every session.
When invoked 972 The same file — it is already loaded in full.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.00972 $0.00972
Opus 5 $0.00486 $0.00486
Sonnet 5 $0.00194 $0.00194
Haiku 4.5 $0.00097 $0.00097

Measured 2d ago against content hash 29ad06f0cfd7, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

Synapsor-Runner 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.

AGENTS.md · 134 lines

How it starts

The opening of the file, as written. The whole thing — 134 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Agent Guide For Synapsor Runner

Synapsor Runner is an OSS MCP safety layer for Postgres/MySQL agents. It replaces raw database tools such as execute_sql with reviewed business capabilities, proposal-first writes, approval outside the model, guarded writeback, and local evidence/replay.

First Commands

Use the stable package for public/user-facing examples:

npx -y @synapsor/runner demo --quick
npx -y @synapsor/runner audit --example dangerous-db-mcp

Use the local checkout while editing this repo:

corepack pnpm install
./bin/synapsor-runner demo --quick --no-interactive
./scripts/verify-release-gate.sh

What Not To Change Casually

Do not weaken these safety boundaries without a failing test and an explicit design note:

  • no model-facing execute_sql, raw SQL, approval, commit, apply, or writeback tools;
  • trusted tenant/principal context must come from backend/session/env config, not model-controlled arguments;
  • proposals must not mutate the source database before approval;
  • direct writeback must enforce tenant, primary key, allowed columns, expected-version/conflict guard, affected-row count, idempotency, and receipt recording;
  • app-owned handlers must re-check tenant/scope, expected version, idempotency, allowed business action, transaction/rollback, and safe error receipts;
  • local store lease guards and stale-lease reclaim must keep concurrent writers from corrupting the SQLite store.

How The Flow Works

MCP tool call
-> trusted context
-> scoped read from Postgres/MySQL
-> evidence/query audit
-> proposal diff
-> approval outside MCP
-> direct writeback or app-owned executor
-> receipt/replay

The source database remains the source of truth. Runner stores local evidence, proposals, receipts, query audit, and replay in SQLite.

Add A Capability Without Reading dist/

Start from schema inspection or a recipe, then run the smoke boundary before wiring an MCP client:

export DATABASE_URL="postgres://readonly:...@localhost:5432/app"
./bin/synapsor-runner onboard db \
  --from-env DATABASE_URL \
  --engine postgres \
  --schema public \
  --table invoices \
  --primary-key id \
  --tenant-column tenant_id \
  --conflict-column updated_at \
  --mode review \
  --visible-columns id,tenant_id,status,late_fee_cents,updated_at \
  --namespace billing \
  --object-name invoice \
  --id-arg invoice_id \
  --patch late_fee_cents=fixed:0 \
  --write-url-env SYNAPSOR_DATABASE_WRITE_URL \
  --yes \
  --output synapsor.runner.json

./bin/synapsor-runner config validate --config ./synapsor.runner.json
./bin/synapsor-runner tools preview --config ./synapsor.runner.json --store ./.synapsor/local.db
./bin/synapsor-runner smoke call --config ./synapsor.runner.json --store ./.synapsor/local.db

Read the full file on GitHub · 134 lines

Changes

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.

  1. 2d ago First seen · 134 lines · 972 tokens per session scan A 29ad06f0cfd7

Subscribe to this mod's changes

Synapsor-Runner AGENTS.md is an instructions file published in the GitHub repository Synapsor/Synapsor-Runner (2 stars, last pushed 5d ago), licensed Apache-2.0. It adds 972 tokens to every session, about $0.0049 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.

Related

Other instructions, from other repositories

AgentAction AGENTS.md

Instructions for dinpd/AgentAction, covering agentaction delivery instructions and release responsibility.

dinpd/AgentAction · 310 tokens

mcp-sentinel CLAUDE.md

Claude Code instructions for kirandevihosur74/mcp-sentinel, covering claude.md, what this is, rules, bright data scraper configuration and registries.

kirandevihosur74/mcp-sentinel · 1,016 tokens

neon-mcp-server CLAUDE.md

Claude Code instructions for voyagi/neon-mcp-server, covering neon mcp server — upwork portfolio project, purpose, what to build, demo scenario: "techstart crm" and value proposition.

voyagi/neon-mcp-server · 1,366 tokens

mcp-database CLAUDE.md

Claude Code instructions for DiegoBulhoes/mcp-database, covering claude.md, commands, architecture, invariants (do not weaken) and adding or changing a tool.

DiegoBulhoes/mcp-database · 2,571 tokens

kasa-mcp AGENTS.md

AGENTS.md instructions for aikadimsoy/kasa-mcp, covering kasa — çalışma kuralları / working rules, ölçüm disiplini (asıl ip — atlanmaz), sessiz arıza — bu projenin baş düşmanı, tehdit modeli — aktör etiketi (zorunlu) and bilinen ortam tuzakları.

aikadimsoy/kasa-mcp · 935 tokens

kasa-mcp CLAUDE.md

Claude Code instructions for aikadimsoy/kasa-mcp, covering kasa — çalışma kuralları / working rules, ölçüm disiplini (asıl ip — atlanmaz), sessiz arıza — bu projenin baş düşmanı, tehdit modeli — aktör etiketi (zorunlu) and bilinen ortam tuzakları.

aikadimsoy/kasa-mcp · 935 tokens