Skip to content

Add ShieldCortex: persistent memory & security for AI agents#386

Open
mkdelta221 wants to merge 1 commit intoanthropics:mainfrom
mkdelta221:add-shieldcortex-skill
Open

Add ShieldCortex: persistent memory & security for AI agents#386
mkdelta221 wants to merge 1 commit intoanthropics:mainfrom
mkdelta221:add-shieldcortex-skill

Conversation

@mkdelta221
Copy link
Copy Markdown

@mkdelta221 mkdelta221 commented Feb 13, 2026

Summary

Adds ShieldCortex — a persistent memory system with built-in security for AI agents.

npm: shieldcortex (6,200+ downloads)
GitHub: Drakon-Systems-Ltd/ShieldCortex (10 stars)
License: MIT

What it does

ShieldCortex gives AI agents a brain that persists between sessions. Every memory write passes through a 6-layer defence pipeline that blocks prompt injection, credential leaks, and memory poisoning attacks.

Key features

  • Persistent memory — SQLite-backed, survives restarts and compaction
  • Semantic search — find memories by meaning, not just keywords
  • Knowledge graph — automatic entity/relationship extraction from memories
  • Memory decay — old, unaccessed memories fade like a real brain
  • Contradiction detection — flags when new memories conflict with existing ones
  • Memory consolidation — merges similar/duplicate memories automatically
  • 6-layer defence pipeline — input sanitisation, pattern detection, semantic analysis, structural validation, behavioural scoring, credential leak detection
  • Skill scanner — detects hidden prompt injection in SKILL.md, .cursorrules, CLAUDE.md
  • Forensic audit trail — full log of every memory operation
  • OpenClaw plugin — real-time LLM input/output scanning via extensions directory
  • Cloud sync — optional fire-and-forget audit sync to ShieldCortex Cloud
  • Python SDKpip install shieldcortex for Python agent frameworks

MCP tools provided

remember, recall, forget, get_context, start_session, end_session, graph_query, graph_entities, graph_explain, consolidate, detect_contradictions, scan_memories, scan_skill, audit_query, quarantine_review, defence_stats, memory_stats, set_project, get_project

Install

npm install -g shieldcortex
shieldcortex install

ShieldCortex gives AI agents persistent memory with knowledge graphs,
decay-based forgetting, contradiction detection, and a 6-layer defence
pipeline that prevents memory poisoning attacks.

MCP server: shieldcortex (npm install -g shieldcortex)

Features:
- Semantic search, project scoping, activation/salience scoring
- Knowledge graph extraction (entities + relationships)
- Memory consolidation and automatic decay
- 6-layer defence pipeline (input sanitisation, pattern detection,
  semantic analysis, structural validation, behavioural scoring,
  credential leak detection)
- Forensic audit trail of all memory operations

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
@mkdelta221
Copy link
Copy Markdown
Author

Quick update since submission:

  • 6,200+ npm downloads (up from 2,300 at submission)
  • 10 GitHub stars
  • New in v2.12: OpenClaw real-time plugin, Python SDK (pip install shieldcortex), cloud sync, version diagnostics
  • Actively maintained — 12 releases in the last week

Happy to make any changes needed to the SKILL.md. Thanks for reviewing!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant