An honest comparison
Backbrief Kit vs. a skill pack
A skill pack gives you procedures: written methods, one per job, that load when the job comes up. That is a real thing to buy and we ship four of them inside the free kit. What a pack cannot give you is anyone to run them, any constraint they have to obey, any check before the output leaves, or any memory of what you decided last week. Here is the trade, including the part where a pack wins.
Page updated August 26, 2026.
The trade, side by side
| Dimension | Procedures A skill pack | Free Backbrief Kit |
|---|---|---|
| What you install | A library of procedures. Each file teaches one method well. | The procedures plus the team that runs them: an orchestrator, specialist agents, always-on rules, a separate verifier, and project state. Plain Markdown, all of it. |
| Who drives | You do. You pick the skill, open the session, and hold the thread. | An orchestrator reads your goal, writes the plan to a file, and routes each unit to the agent that owns that domain. |
| Portability | The real advantage, and it is real. A plain procedure carries to whichever assistant you are using. | Built for Claude Code. Subagents, model tiers, permission rules, and slash commands are the reason every row below this one exists, and none of them port. |
| Constraints | None available. A procedure cannot stop a session doing something. | Rules every agent obeys whether or not it remembered to: no spend, no send, no publish without your go. Drafts wait in a folder until you say the word. |
| Who checks the work | The session that wrote it. It says the work is done, because it always does. | A separate verifier reads the finished artifact in fresh context, holding the acceptance criteria and nothing else. Not the draft history, not the reasoning. |
| Cost discipline | Every step runs on whatever model the session is on. | A written model ladder. Cheap models take mechanical work with pass/fail criteria; the strong model is kept for judgment and for verification. |
| Memory between sessions | Whatever you paste back in. | Plans, decisions, and handoff briefs live in files. /handoff writes the brief, /pickup reads it, and an opt-in hook recalls it at session start. |
| Standardizing a team | Everyone gets the same procedures. Everyone still runs them their own way. | Everyone installs the same agents, rules, and gates from one repo. /verify-install reports the version in force and anything local shadowing it. |
| Changing it | Markdown you can edit. | Markdown you can edit. Installs never overwrite files you already have. |
| Cost | Varies by pack. | Free. Business OS ($79 one-time) adds the business layer on top, not a toll on the kit. |
Where a skill pack wins, and it is not close
A good pack runs anywhere. The same Markdown loads in Claude, in Cursor, in whatever you open next year, because a procedure makes no assumptions about the tool reading it. If you move between assistants during a normal week, that portability beats anything on our side of the table, and you should buy the pack.
We went the other way on purpose. Backbrief runs in Claude Code and uses what Claude Code gives us: subagents with their own context, model tiers, permission rules the engine enforces, commands you type. Agents, rules, gates and cost-tiering exist because of those. A pack built to load in four tools can only ship what all four share, and what they share is the procedure.
Depth in one tool or reach across four. That is the actual choice, and neither answer is embarrassing.
Which one you need
Buy the packif you want one method done well, you switch assistants often, or you already have a setup you like and want to drop a procedure into it. Packs are cheap, portable, and they do not ask you to adopt anyone’s opinions about how work should move.
Install the kit if the failure you keep hitting is coordination rather than method. If the thread unravels past the hundredth message. If work goes out without a second pair of eyes. If a new chat means re-explaining the project from scratch. If two people on one repo are getting two different answers from the same tool.
Doing both is normal. The kit ships four skills and reads any others you install. Nothing in it stops you adding a pack tomorrow. The layers are the point, and skills are one of them.
A skill is the smallest unit. Backbrief ships the system that runs them.
Skills. The procedure. A written method for one job, loaded when the job comes up.
Agents. Someone whose job is to run it. Each owns a domain, sits at a model tier, and stays inside a scope.
Rules. The constraints every agent obeys whether or not it remembered to.
Gates. A separate verifier reads the finished artifact in fresh context before it ships.
State. Plans, decisions, and handoff briefs live in files, so a new chat picks up where the last one stopped.
Install it and read the Markdown.
Ten minutes, one install and a /setup interview. Every file is plain Markdown you can open, edit, or ignore. If it is not for you, the read is still worth it.
Weighing this against building your own setup instead? Read that comparison.