Agent skillsMemory & context

ECC

affaan-m/ECC

A performance layer for agent harnesses — skills, instincts, memory, and security bundled into one system.

Repository
Stars
239k
Forks
36.3k
Open issues
139
Language
JavaScript
Licence
MIT
Created
2026-01-18
Last push
2026-08-10
Effort to adopt

Some wiring

A few days of integration work — credentials, data mapping, a deploy target.

01

What it actually does

ECC positions itself as an optimisation system that sits on top of whatever harness you already use — Claude Code, Codex, OpenCode, Cursor. Rather than replacing the agent, it improves how it performs.

The bundle covers four areas: skills (reusable capability packs), instincts (default behaviours that fire without being asked), memory (context that survives sessions), and security. It also pushes research-first development, meaning the agent gathers information before acting rather than after being corrected.

It ships as npm packages (`ecc-universal`, `ecc-agentshield`), a GitHub App, and a plugin, and is translated into twelve languages.

One thing worth flagging because the maintainers flag it themselves: the README opens with a prominent warning to install only from official channels, because third-party mirrors of a popular agent tool are an obvious malware vector. Take that seriously — verify the source before installing anything that will hold your credentials and read your codebase.

02

Who it's for

  • 01

    Teams already invested in an agent harness who want more out of it

  • 02

    Anyone who wants memory and security handled as one system rather than three integrations

  • 03

    Developers who keep correcting an agent that acted before it researched

  • 04

    Non-English teams — the documentation is genuinely multilingual

EngineeringOperationsSaaSAny industry
03

Where it earns its keep

  • Adding persistent memory to a harness that does not have it natively
  • Applying a consistent security posture across every agent on the team
  • Encoding team defaults as instincts so nobody has to prompt for them
  • Getting research-before-action behaviour without writing the prompt yourself
  • Standardising agent behaviour across a team using mixed tooling
04

Use it, or skip it

Reach for it when

  • Your harness works but its defaults are not your team's defaults
  • You want memory, skills, and security from one source rather than assembled
  • Agents on your team act too fast and research too little
  • You need the setup documented in a language your team actually reads

Skip it when

  • You are cautious about layering third-party code into your agent's execution path — that is a reasonable position here
  • Your harness already provides these capabilities natively
  • You want minimal dependencies; this is an opinionated bundle
  • You cannot verify the install source, in which case do not install it at all
05

10 automations

  1. 01Operations

    Team-wide agent defaults

    Encode your engineering standards as instincts so every agent applies them without anyone remembering to prompt.

  2. 02Engineering

    Persistent project memory

    Give agents recall of past decisions on a project so they stop proposing approaches that were already rejected.

  3. 03Operations

    Security posture enforcement

    Apply consistent guardrails on what agents may read or execute across every machine on the team.

  4. 04Engineering

    Research-first requirement

    Force the agent to gather context on unfamiliar code before proposing a change, cutting confidently wrong edits.

  5. 05Operations

    Multilingual team rollout

    Roll out identical agent capability to teams working in different languages using the translated documentation.

  6. 06Operations

    Skill library curation

    Maintain an approved internal skill set and distribute it rather than each engineer installing ad hoc.

  7. 07Engineering

    Supply-chain verification job

    Automate a check that installed agent tooling came from the official package and matches expected hashes.

  8. 08Engineering

    Cross-harness consistency

    Keep behaviour comparable whether an engineer uses Claude Code, Cursor, or Codex.

  9. 09Operations

    Onboarding capability bundle

    Install one package on a new hire's machine and have their agent match the team's on day one.

  10. 10Engineering

    Behaviour regression checks

    Run a standard task set after each update to confirm agent behaviour has not silently changed.

Want one of these running by Friday?

LimeDock builds these as real workflows inside your stack — deployed to your cloud, wired into your Slack and CRM, with the code in your repo. You pay a build fee and your own API keys, nothing else.

Book a workflow call
07

Source

Repository stats were read from the GitHub API and reflect the last time we refreshed this entry. The editorial breakdown above is LimeDock’s own analysis — we are not affiliated with affaan-m.

https://github.com/affaan-m/ECC