- Home
- Skills
- Workflow Automation
- Disciplined Agent Pack
Works with the AI Agents you already use
Disciplined Agent Pack
A full discipline system for AI coding agents: a design doc you approve before code, project memory that stays current, and append-only safety for index files.
Free
Disciplined Agent Pack
Example session with this skill installed
Add full-text search to my markdown notes CLI. Around 10k notes, must feel instant.
- Read your context and instructions
- Compiled the disciplined agent
Design doc: full-text search (awaiting your approval)
Problem. Search scans every note on every query; at 10k files that is seconds of waiting.
Options.
| Latency | Disk cost | Complexity | |
|---|---|---|---|
| 1. Brute scan + mtime cache | ok at 10k | none | low |
| 2. Token → file index | instant | ~2 MB | medium |
| 3. Embedded search library | instant | heavy | high |
Decision: option 2. Build the index on startup (mtime_ns + size stats), rebuild a single file when it changes. Query routing: valid token → index, anything else → scan.
Migration check. Existing index files are a subset of whole-word tokens; any zero-hit query falls back to scan, so stale results cannot leak.
Status: awaiting user approval — no code has been written yet.
`
Connects securely to your tools. The creator never sees your data.
About this skill
Four habits that keep an AI coding agent useful on a real project: design first, remember everything, never corrupt ordered files. One skill, three moving parts that fire at the right moments.
What is inside
- Feature Architect — every non-trivial feature starts with a short design doc (problem, 2-3 options with trade-offs, data flow, risks) that you approve before a single line is written. Hard stop until approval; trivial fixes skip the gate.
- Update Memory — after each completed work block the agent syncs project memory: what changed, why it was decided that way, which pitfalls were hit. Four templates included (current state, code map, architecture, patterns) plus a pruning rule so memory never bloats.
- Append-Only Index — a tiny Node CLI for index/manifest files where line order defines identity: safe appends, duplicate protection after path normalization, and a git-based --verify that proves the file only gained lines. 35-scenario test battery included.
Why it works
Each part covers the others' blind spots: the gate prevents chaos, memory prevents re-litigated decisions, the index tool prevents silent ID corruption. Born in daily work on a real Java voxel engine, not in theory.
Node.js >= 10.12 is needed only for the index utility; everything else has no runtime requirements.
Requires Node.js 10.12+ for append-only operations. Verification features require Git on PATH. Memory updates require manual user confirmation of working code.
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
Fresh listing
Recently published to Agensi
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 22 days ago
- Free with an Agensi account
Requires