Skill Detail

Run local small-model coding sessions with SmallCode

Use SmallCode as a terminal-native coding agent for 8B-35B local models, with budget-managed context, forgiving tool parsing, TODO-driven planning, and patch-based edits.

Developer ToolsCustom Agents
Developer Tools Custom Agents Published
⭐ 967 GitHub stars ⬇ 7.5k/wk npm
COPY SKILL INSTRUCTIONS (OPTIONAL)
npx skills add agentskillexchange/skills --skill run-local-small-model-coding-sessions-with-smallcode Copy
Uses the third-party skills CLI, not an ASE-owned installer. Check your agent’s compatibility. This copies instructions; complete the upstream tool setup below separately.
At a glance
Tools required
Node.js/npm, a local 8B-35B coding model or compatible local model runtime, terminal access to the target project
Install & setup
Install globally with npm install -g smallcode, or run directly with npx smallcode, then start it from the target project directory.
Author
Doorman11991
Publisher
Individual
Last updated
May 21, 2026
Quick brief

Use SmallCode when an operator wants a coding agent that can get useful work from local 8B-35B models on consumer hardware instead of assuming frontier-model context windows and reliable tool calling. The agent/operator installs the npm package, runs SmallCode in a project directory, and uses its budget-managed context, summarized state, forgiving multi-format tool parser, TODO-file planning, and search-and-replace patch editing to carry out coding tasks. Keep the scope to local small-model coding-agent sessions; this is not a generic benchmark card, model leaderboard, or broad developer-tools listing.

How it works

What this skill actually does

Inputs and prerequisites: Node.js/npm, a local 8B-35B coding model or compatible local model runtime, terminal access to the target project.

Setup notes: Install globally with npm install -g smallcode, or run directly with npx smallcode, then start it from the target project directory.

Source and verification boundary: use https://github.com/Doorman11991/smallcode as the canonical reference before running the workflow; keep commands, API calls, CLI usage, and generated outputs reviewable against that upstream source.

Framework fit: publish this as a Custom Agents workflow only when the operator can invoke the documented toolchain directly, rather than treating the upstream project as a generic product listing.