gsc-cli

module
v0.0.0-...-39bfe36 Latest Latest
Warning

This package is not in the latest version of its module.

Go to latest
Published: Jul 18, 2026 License: Apache-2.0

README

GitSense: CLI

Use gsc to record what matters and make it available to coding agents on demand.

GitSense captures knowledge while you work or extracts it at scale with GitSense Chat, then delivers relevant files, lessons, and rules to coding agents through gsc.

gsc is the left and bottom of the picture. It records knowledge while you work and gives agents a simple way to find it later. GitSense Chat is optional. It adds the right side when you need to generate and maintain knowledge across one repository or many.

Why gsc?

Useful knowledge comes up while you are building, debugging, reviewing, and planning. gsc gives you and your agents a simple place to keep it, enough structure to find it again, and a way to pull in only what matters for the work at hand.

gsc makes knowledge easier to record, find, and use when an agent needs it.

  • Record it. Save notes, lessons, and rules for yourself or with the repository.
  • Find it. Narrow the knowledge by topic, tag, summary, or related file.
  • Use it when it matters. Give the agent the few relevant records instead of loading everything into every conversation.
Router for Skills

gsc does not replace the skills you have already built. It acts as a router in front of them. Ask your agent to review your skills and create notes describing when each one should be used and where the useful parts live. Later, the agent can query those notes before deciding which skill or section to load. The skill stays the source of truth, but its full contents stay out of context until the work calls for them.

For a small, stable set of skills, a table of contents or an if/else Markdown file may be enough. As that file grows, the agent still has to load it and reason through every branch. With gsc, the agent can query the routing records and inspect only the matches.

The same approach works for AGENTS.md, CLAUDE.md, runbooks, and other instructions that should remain the source of truth.

Record What Matters

Most of the time, you will not write a gsc command yourself. Tell your agent what should be remembered:

Record a repository note that cache entries are scoped by tenant.
Attach it to internal/cache/store.go.

When something goes wrong or takes too long:

Save a repository lesson that migration tests must run after schema changes
because the unit tests do not cover rollback behavior.

When guidance should be checked during the work:

Add a repository rule that blocks rm -rf commands.

After gsc experts init, the agent knows how to turn those requests into structured records. It uses the same commands and record schema every time. gsc validates the record, saves it at personal or repository scope, and gives it a stable identity so it can be found and updated later.

This removes the guesswork around where new knowledge belongs. Repository records can be committed with the code. Personal records follow you without becoming part of the repository.

Find It Later

Search everything together:

gsc knowledge search "cache boundaries"

Or ask for knowledge attached to a file:

gsc notes get --file internal/cache/store.go
gsc rules get --file internal/cache/store.go

An integration can also decide when rules are checked. pi-brains does this for the Pi coding agent, so important guidance does not depend on the agent remembering to look for it.

Quick Start

Review the install script, then install gsc:

curl https://raw.githubusercontent.com/gitsense/chat/refs/heads/main/install.sh | bash

From inside a repository, ask your coding agent:

Run `gsc experts init` and follow the instructions it prints before you start working in this repository.

See the usage guide for scopes, knowledge discovery, Brains, enriched search, sessions, GitSense Chat commands, and building from source.

License

See LICENSE.

Jump to

Keyboard shortcuts

? : This menu
/ : Search site
f or F : Jump to
y or Y : Canonical URL