accessflow CLAUDE.md

Project instructions for AccessFlow, an open-source platform that controls access to databases by reviewing and approving queries before they run. It also covers permissions, audit records, and notifications.

In plain words
What is it for?
Use them when developing AccessFlow’s database proxy, query approval workflows, access roles, audit logging, or related backend code.
Why use it?
They tell coding agents which project patterns and design documents to follow, helping changes fit AccessFlow’s security-sensitive architecture.

Instructions file

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/bablsoft/accessflow/claude-md
Clone the repo
git clone --depth 1 https://github.com/bablsoft/accessflow
Per session 12,646 This file is loaded in full into every session.
When invoked 12,646 The same file — it is already loaded in full.
Security scan B 2 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 $0.12646 $0.12646
Opus 5 $0.06323 $0.06323
Sonnet 5 $0.02529 $0.02529
Haiku 4.5 $0.01265 $0.01265

Measured 2d ago against content hash 1089e0458484, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade B, and why

accessflow CLAUDE.md scanned grade B with 2 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.

Reads agent configuration directoriesmediumAgent snooping

.claude/, .codex/, .gemini/ hold keys, settings and other credentials a mod has no legitimate need for.

`default`, so the compiler will not find them all → `.claude/patterns/notification-fanout.md`

Makes network callslowCapability

Not a fault in itself. Listed so you know the mod talks to something, and to what.

├── deploygov/ # Deployment approval governance (epic AF-682): gate CI/CD deployments behind approval workflows. #684 landed the persistence foundation — pipelines/environments/freeze windows/requests/decisions/routing/g
CLAUDE.md · 703 lines

How it starts

The opening of the file, as written. The whole thing — 703 lines — stays where its author put it; the contents beside it link to each section on GitHub.

CLAUDE.md — AccessFlow Agent Rules

This file is the authoritative guide for AI agents implementing AccessFlow. Read it entirely before writing any code. When in doubt, prefer the rules here over general best-practice intuition.

Before writing code in an area, open .claude/patterns/README.md and read the matching pattern. This file holds the rules that are always true; the patterns hold the per-area detail — canonical examples with file:line, acceptance checklists, and the anti-patterns whose failure mode is silent. docs/ remains the design reference for what to build; the patterns describe how.


Project at a Glance

AccessFlow is an open-source database access governance platform. It is a full query proxy between users and customer databases, enforcing configurable review and approval workflows before any query executes. Core capabilities: AI-powered query analysis, multi-stage human approval chains, role-based access control, a tamper-evident audit log, and real-time notifications.

Supported engines. The relational engines (PostgreSQL, MySQL, MariaDB, Oracle, SQL Server) run over pooled JDBC in-process; any other JDBC engine can be added by uploading its driver JAR (db_type=CUSTOM). Everything else is an engine plugin — a standalone shaded JAR implementing the core.api.QueryEngine SPI, resolved on demand through the connector catalog, SHA-256 verified, and discovered via ServiceLoader:

Category Engines
RELATIONAL PostgreSQL, MySQL, MariaDB, Oracle, SQL Server, CUSTOM (in-process JDBC)
WAREHOUSE Snowflake, BigQuery, Databricks
DOCUMENT MongoDB, Couchbase
KEY_VALUE Redis, DynamoDB
WIDE_COLUMN Cassandra, ScyllaDB (same JAR)
SEARCH Elasticsearch, OpenSearch (same JAR)
GRAPH Neo4j

Each plugin has its own version line, pinned by URL + SHA-256 in connectors/<id>/connector.json; CI fails on pin drift. Per-engine specifics — dialect classifier, row-security shape and where it fails closed, connection model, shading — live in docs/05-backend.md (one section per engine). The host↔plugin contract and the add-an-engine checklist are docs/15-engine-sdk.md.

Read the full file on GitHub · 703 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 · 703 lines · 12,646 tokens per session scan B 1089e0458484

Subscribe to this mod's changes

accessflow CLAUDE.md is an instructions file published in the GitHub repository bablsoft/accessflow (4 stars, last pushed 2d ago), licensed Apache-2.0. It adds 12,646 tokens to every session, about $0.0632 per session on Opus 5. A static security scan graded it B with 2 findings (reads agent configuration directories, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.