From composto
Generates health-aware intermediate representations (IR) for any code file at L0-L3 levels: structure map, compressed code+health signals, deltas, raw source. Saves tokens for understanding or sharing context.
npx claudepluginhub mertcanaltin/composto --plugin compostoThis skill uses the workspace's default tool permissions.
Generate compressed, health-annotated code representation. Send meaning, not code.
Initializes Composto proactive monitoring, Health-Aware IR, and watcher engine for automatic detection of security issues, code smells, and codebase health trends at session start.
Provides deep explanations of specific files or modules, including why structured this way, git history, contributors, refactors, complexity, and lenskit metrics/graph context.
Share bugs, ideas, or general feedback.
Generate compressed, health-annotated code representation. Send meaning, not code.
# L0: Structure Map (~10 tokens) — just file outline
npx composto ir <file> L0
# L1: Health-Aware Generic IR (~85 tokens) — compressed code + health
npx composto ir <file> L1
# L2: Delta Context (~65 tokens) — only what changed + health
npx composto ir <file> L2
# L3: Raw Source — original code (fallback)
npx composto ir <file> L3
| Need | Layer | Tokens |
|---|---|---|
| "What's in this file?" | L0 | ~10 |
| "What does this file do?" | L1 | ~85 |
| "What changed recently?" | L2 | ~65 |
| "Show me the exact code" | L3 | variable |
src/auth/session.ts
FN:createSession L5
FN:validateToken L23
CLASS:SessionManager L45
USE:jsonwebtoken{sign,verify}
FN:createSession({credentials}) [HOT:12/30 FIX:67% COV:↓ INCON]
VAR:token = sign(credentials, secret)
RET {token, expiresAt}
Health annotations (only on unhealthy code):
[HOT:12/30] — 12 changes in last 30 commits (hotspot)[FIX:67%] — 67% of changes were bug fixes[COV:↓] — test coverage declining[INCON] — inconsistent patterns from multiple authorsRaw source code wastes tokens on syntax, indentation, and boilerplate that LLMs already know. Health-Aware IR strips the noise and adds health signals that raw source never had. Less tokens, more insight.