- Home
- Skills
- Productivity
- infernoflow: git-native memory for coding agents
Works with the AI Agents you already use
infernoflow: git-native memory for coding agents
Memory for coding agents: gotchas, decisions, dead ends, bookmarks and session handoffs, saved as plain text in your git repo.
Free
infernoflow: git-native memory for coding agents
Example session with this skill installed
Add error handling to the login call in src/api/auth.js
- Read your context and instructions
- Compiled the infernoflow
Summary
Loaded project memory first. There's a gotcha logged for this file, so I didn't rely on response.ok.
🔥 (from infernoflow memory) gotcha at src/api/auth.js:42: the auth API returns 200 even on errors. Check the response body, not the status code.
export async function login(email, password) {
const res = await fetch("/api/auth/login", {
method: "POST",
headers: { "Content-Type": "application/json" },
body: JSON.stringify({ email, password }),
});
const data = await res.json();
// The auth API returns 200 on failures too (see memory), so trust the body.
if (!data.ok) {
throw new AuthError(data.error?.code ?? "LOGIN_FAILED", data.error?.message);
}
return data.user;
}
| Memory used | Type | Where |
|---|---|---|
| Auth API returns 200 on errors | gotcha | src/api/auth.js:42 |
| Opaque session tokens, not JWTs, so sessions can be revoked | decision | src/auth.js |
Logged to memory
- decision:
login()throwsAuthErrorwith the server's error code so the UI can map it to a message (src/api/auth.js)
Next steps
- Map the
AuthErrorcodes to messages in the login form - Run
infernoflow statusto see the new entry
Connects securely to your tools. The creator never sees your data.
About this skill
You come back to a repo after a week and spend an hour working out why you didn't use that library, or how you fixed that weird build error last time. Your agent has it worse: it starts every session from zero. This skill gives it a memory that lives inside your repo, using infernoflow.
What it does
- Logs gotchas: undocumented behavior and config quirks that cost you time
- Records decisions together with the reason behind them
- Tracks dead ends, so nobody tries the same failed fix twice
- Remembers lasting preferences across sessions
- Drops bookmarks: named resume points with where you stopped and what's next
- Writes a handoff for the next session, so you, a teammate or another agent start with context
- Keeps working when the agent forgets to log: prompt hooks catch moments like "still broken" and record them
How it works
- The first time, the skill asks before installing anything, then runs
npm install -g infernoflowandinfernoflow init --yes. That creates the.ai-memory/store, the rule files and the MCP config. - At the start of a session the agent calls
amp_resumeto load recent bookmarks, decisions and open dead ends. - While you work it logs with
amp_write(gotcha, decision, attempt, preference). - Before trying something, it checks memory with
amp_readoramp_search.
What makes it different
- The memory is plain JSONL in
.ai-memory/, committed with your code. You can read it, diff it and review it in a PR. - It travels with your branch. A teammate who checks out your branch gets your findings too.
- No server, no database, no account.
- The same memory works in Claude Code, Cursor and GitHub Copilot.
Frameworks & tools
Needs the free infernoflow npm package (Node 18+). Works with Claude Code, Cursor and VS Code Copilot through MCP, or with any agent through the CLI. macOS, Linux and Windows.
Use cases
- Picking up a big refactor after a weekend without losing the thread
- Bringing a new teammate or agent up to speed on the repo's tribal knowledge
- Stopping the agent from suggesting a library you already tried and dropped
- Keeping architectural decisions and the reasons behind them in one place
Privacy
The npm install is the only network call. infernoflow makes none after that, and telemetry stays off unless you turn it on. Secrets are redacted before anything is written.
Free and open source (MIT).
Website: https://infernoflow.dev · npm: https://www.npmjs.com/package/infernoflow · GitHub: https://github.com/ronmiz/infernoflow · VS Code extension: https://marketplace.visualstudio.com/items?itemName=infernoflow.infernoflow
Requires Node 18+ and a one-time global npm install. Memory entries can become stale if referenced files are modified. Shared via git, so team members must pull to see updates.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
2 uses
Times used to date
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 1 month ago
- Free with an Agensi account
Requires
Permissions
Allowed Hosts
File Scopes
Shell: runs the infernoflow CLI (resume, ask, log, bookmark). Network: only the one-time npm install, and only after the user agrees. infernoflow makes no network calls after that, and telemetry is off unless the user opts in. Files: memory goes to .ai-memory/, and a marked block is added to the rule files. During setup, init also writes the MCP configs (adding .mcp.json to .gitignore) and installs this skill and the memory-keeper agent into .claude/.