What is prism-mem CLAUDE.md?

Project instructions for a system that builds a knowledge graph from coding-session transcripts and Git history. A knowledge graph stores facts and relationships, while vector embeddings help find similar entities or concepts.

Use them when working on extracting facts, linking entities, scoring relationships, or storing and searching the resulting knowledge graph.

They document the project’s technology choices, processing stages, storage, and current architecture so changes fit the existing design.

How rahult18/prism-mem configures Claude Code

Reuse

Borrowing it

Nothing to install: this file belongs to rahult18/prism-mem. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.

Copy the file

curl -O https://raw.githubusercontent.com/rahult18/prism-mem/main/CLAUDE.md

Clone the repo

git clone --depth 1 https://github.com/rahult18/prism-mem

Made for: Claude Code.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

README.md

[![agentmods](https://agentmods.dev/badge/instructions/rahult18/prism-mem/claude-md.svg)](https://agentmods.dev/instructions/rahult18/prism-mem/claude-md)

Your own site

<a href="https://agentmods.dev/instructions/rahult18/prism-mem/claude-md"><img src="https://agentmods.dev/badge/instructions/rahult18/prism-mem/claude-md.svg" alt="Measured on agentmods" height="20"></a>

Per session 727 This file is loaded in full into every session.

When invoked 727 The same file — it is already loaded in full.

Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.

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.

Model	Per session	Once invoked
Fable 5.1	$0.00727	$0.00727
Opus 5	$0.00364	$0.00364
Sonnet 5	$0.00145	$0.00145
Haiku 4.5	$0.00073	$0.00073

Measured 6d ago against content hash 6095bd2c1360, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-06, from the pricing page.

Security

Grade A, and why

prism-mem CLAUDE.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 6d 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.

CLAUDE.md · 44 lines

How it starts

The opening of the file, as written. The whole thing — 44 lines — stays where its author put it; the contents beside it link to each section on GitHub.

CLAUDE.md

Overview

This project implements prism-mem, a knowledge graph generation system that extracts structured facts from coding sessions and git history, storing them in a SQLite database with vector embeddings. Claude processes session transcripts and code diffs through multiple phases (extracting relations, linking entities, scoring triples, and generating constitutions), using the Anthropic API via LiteLLM to power the extraction pipeline while maintaining session state across subagent calls.

Tech Stack

Anthropic API (Haiku) — Powers the extraction phase; accessed via LiteLLM for standardized API calls
SQLite with sqlite-vec extension — Persists the knowledge graph and enables vector similarity searches for entity linking
Transformer models — Generate embeddings for entity linking using similarity thresholds
Python — Core implementation language with session readers and git diff analyzers

Architecture

The system flows through distinct phases:

Phase 2 (Completed) extracts initial relations from transcripts and code diffs via the extractor, which calls the Haiku API and produces Graph.relations
Phase 3 (current) performs entity linking by calling link_triple() which invokes create_edge(), using embeddings and similarity thresholds to connect related entities
Phase 4 (next) continues processing
Phase 7 scores triples and runs the constitution generator

Key Components:

session_reader.py — Extracts sessionId from transcripts stored in ~/.claude/history.jsonl
git_reader.py — Implements read_git_diff() for code change extraction
mcp_server.py — Server component handling tool execution
Subagents — Generate their own .jsonl session files and produce tool_result objects
~/.claude/projects/ — Stores project-specific .jsonl files containing knowledge graph data

Key Decisions

Haiku via LiteLLM — Chose Haiku (not more expensive models) for cost-efficient extraction, accessed through LiteLLM for unified API handling
Vector embeddings for linking — Entity linking uses transformer-based embeddings with configurable similarity thresholds rather than exact string matching, enabling fuzzy entity resolution
Multi-phase architecture — Separated extraction (Phase 2), linking (Phase 3), scoring (Phase 7), and constitution generation (Phase 7) into distinct phases rather than single-pass processing, enabling iterative refinement
Session-per-subagent pattern — Each subagent maintains its own .jsonl session file rather than shared state, providing isolation and auditability
ANTHROPIC_API_KEY authentication — Direct API key management for Anthropic rather than OAuth, enabling programmatic automation

Read the full file on GitHub · 44 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.

6d ago First seen · 44 lines · 727 tokens per session scan A 6095bd2c1360

Subscribe to this mod's changes

prism-mem CLAUDE.md is an instructions file published in the GitHub repository rahult18/prism-mem (2 stars, last pushed 3mo ago), licensed MIT. It adds 727 tokens to every session, about $0.0036 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.

Memory and context

next.js AGENTS.md

AGENTS.md instructions for vercel/next.js, covering next.js development guide, codebase structure, monorepo overview, core package: packages/next and other important packages.

vercel/next.js · 7,296 tokens

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.

openai/codex · 5,182 tokens

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).

microsoft/vscode · 6,785 tokens

prism-mem: Instructions file for Claude Code