/improve

The /improve slash command - implement a focused area, or run a read-only senior audit that writes execution-ready plans.

/improve is a slash command enigma deploys into your coding agents. It improves the current project in one of two modes - it either edits a focused area directly, or audits the codebase read-only and writes execution-ready plans for another agent to run.

It is deployed to Claude Code, OpenAI Codex and opencode (authored once, copied into each app’s command/prompt directory). Run it inside the agent as /improve <args>. The mode is resolved from the arguments.

Looking for the CLI subcommands (enigma install, enigma autoskills, …)? Those are in Commands. /improve runs inside the agent, not the enigma binary - though enigma improve --help prints the same reference.

Two modes

Mode What it does Touches your code?
Implement Edits a focused area to improve it, then runs build/lint/tests and reports. Yes
Advisor Audits the repo like a senior advisor and writes self-contained plans under plans/. Never edits source. No

The mode is chosen from the arguments: an advisor keyword anywhere picks Advisor; otherwise a first-token area picks Implement; anything else prints both usages and stops (it never guesses).

Implement mode

/improve <area> - makes the smallest change that achieves the improvement, applies it incrementally, then verifies (a UI change by looking at the screen, a behavior change by exercising the path - a green build alone is not proof) and reports what changed and why.

Area Aliases Focus
ui frontend Visual design, components, states, accessibility, responsiveness.
security Secrets, authn/authz, input validation, OWASP Top 10, transport/crypto, dependency audit.
performance perf Measure first, then N+1s, indexes, caching, bundle size, payloads - re-measured.
seo Metadata, semantic HTML, structured data, crawlability, Core Web Vitals.
refactor refactorize DRY/consistency pass: one concept, one implementation - no behavior change. 
/improve ui
/improve security
/improve refactor

Advisor mode (read-only)

The advisor never modifies source. It maps the repo, audits across nine categories (correctness, security, performance, tests, tech debt, dependencies, DX, docs, direction), vets every finding against the actual code, then writes self-contained plans a less-capable executor can follow.

Variant What it does
audit [focus] The full workflow. With a focus (e.g. audit security) it audits only that category.
quick / deep Effort level for the audit - composes with anything (quick perf, deep security).
branch Audit only the current branch’s changes vs the default branch.
next (features, roadmap) Forward-looking: grounded direction/feature suggestions, each citing repo evidence.
plan <description> Skip the audit; write a single plan for exactly what you describe.
review-plan <file> Critique and tighten an existing plan in plans/.
execute <plan> Dispatch a cheaper executor on one plan (isolated worktree), then review the diff like a tech lead.
reconcile Process what changed since last session: verify done, refresh drifted, retire dead plans.
--issues Modifier on any planning invocation: also publish each plan as a GitHub issue via gh. 
/improve audit
/improve quick security
/improve next
/improve plan add rate limiting to the public API
/improve execute 003

Plans land in plans/ (a README.md index plus one numbered file per plan), each stamped with the commit it was written against for drift detection, and each written to be executed from the plan file and the repo alone - no memory of the audit required.

Disambiguation

A bare security or performance runs Implement (it edits code - backward compatible). To audit those read-only instead, prefix an advisor keyword: audit security, quick perf, deep security.

From the CLI

/improve runs inside the agent, but the same reference is one command away in your terminal:

enigma improve --help