Skip to content

lerim skill

Install Lerim's bundled skill, register instruction artifacts, and review evidence-backed skill update proposals.

For the end-to-end dashboard review workflow, see Skill Updates.

Install

lerim skill install

This copies the repo-shipped Lerim skill files into standard agent skill locations so supported agents can query Lerim before they start work.

Register Targets

Register any skill or instruction artifact Lerim should monitor:

lerim skill target add ~/.agents/skills/clean-code --description "Keep simplification guidance current"
lerim skill target list
lerim skill target show it_abc123

Supported targets include skill directories, SKILL.md, AGENTS.md, CLAUDE.md, GEMINI.md, and related instruction files. Proposal paths are limited to files scanned as part of the registered target; skill bundles may also propose new files under references/, reference/, or examples/. Targets inside a registered Lerim project are scoped to that project. Targets outside registered projects are global and may learn from registered project records.

Auto-Apply

Targets default to review mode. Auto-apply is opt-in and remains bounded by the target policy:

lerim skill target auto-apply it_abc123 --enable --risk low
lerim skill target auto-apply it_abc123 --disable

Automatic application only runs for pending proposals that pass validation, guard checks, risk limits, changed-file limits, added-line limits, and removed-line limits. Scripts, assets, config files, and frontmatter stay blocked unless the target policy explicitly allows them.

Refresh And Review

Run a scan to compile proposals from scoped context records:

lerim skill refresh it_abc123
lerim skill refresh clean-code --record-limit 120 --json

Review proposals before applying or rejecting them:

lerim skill proposal list
lerim skill proposal list --target-id it_abc123 --status pending_review
lerim skill proposal show ip_abc123
lerim skill proposal apply ip_abc123
lerim skill proposal reject ip_abc123

Applying is allowed only for pending, guard-accepted, validation-passing proposals whose file baselines still match the current target files. Terminal proposals such as applied or rejected proposals cannot be edited back into review.

The dashboard Skills tab shows the same proposal lifecycle with a line-numbered diff, full-file preview, and confirmation flow before writes happen.

What to do next

Add a short Lerim startup instruction to the repository's agent instruction file, such as AGENTS.md, CLAUDE.md, or another tool-specific instruction file. The instruction should tell agents to use Lerim when past project context may matter, not as a ritual for every chat.

The recommended startup flow is:

lerim context-brief show
lerim working-memory show

See Agent Startup Context for the full install, instruction snippet, and verification flow.