Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Lossless session context compression for token efficiency. Extracts entities, decisions, and state into compact format before context window fills. 10-30x reduction in context size while preserving all actionable information.
.claude/skills/session-compression/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-19 | ✗→✓ | ▲ Improved | — | — |
| case-12 | ✗→✓ | ▲ Improved | — | — |
| case-10 | ✗→✓ | ▲ Improved | — | — |
| case-08 | ✗→✓ | ▲ Improved | — | — |
| case-18 | ✗→✓ | ▲ Improved | — | — |
Compress session context without losing actionable information. Use before context window fills up or when handing off between sessions.
| Metric | Raw Context | Compressed | |--------|------------|------------| | Token count | ~50K | ~3-5K | | Decision count | Scattered | Indexed | | File references | Buried in prose | Listed | | Error history | Verbose stack traces | Root cause + fix | | Reduction | 1x | 10-30x |
markdown## A: Actions Completed - [x] Refactored auth middleware (auth.ts, middleware.ts) - [x] Added rate limiting (api/limiter.ts) - [ ] Pending: Write tests for rate limiter ## C: Context - Project: my-app (Next.js 16, TypeScript) - Branch: feat/auth-refactor (3 commits ahead of main) - Blockers: None ## D: Decisions - D1: JWT over sessions (stateless, mobile-friendly) - D2: 100 req/min rate limit (based on load test) - D3: Rejected Redis session store (overkill for current scale) ## E: Entities - Files: auth.ts:42, middleware.ts:15, api/limiter.ts (new) - Deps: jsonwebtoken@9.0.0, express-rate-limit@7.0.0 - APIs: POST /auth/login, POST /auth/refresh, GET /auth/me
Replace verbose references with short codes:
"the authentication middleware in src/middleware/auth.ts" -> "auth.ts"
"the user requested that we use JWT tokens" -> "D1: JWT"
"the PostgreSQL database running on port 5432" -> "pg:5432"The session-compressor hook automatically generates ACDE format before context compression:
Compressed sessions feed into the palace:
Compass uses ACDE format for "where were we?" recovery:
Last session compressed:
A: 3 done, 1 pending (rate limiter tests)
D: JWT auth, 100rpm limit
E: auth.ts, middleware.ts, limiter.ts| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-11 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-04 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-02 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-09 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-06 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-19 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-12 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-13 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-22 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-16 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-10 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-07 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-08 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-18 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-03 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-05 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-15 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-20 | pass→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-17 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-01 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-14 | pass→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-21 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +36 percentage points is the difference between those two pass rates over the 22 comparable cases.
The per-case answers from this run were removed by the retention sweep, so the case table below shows the verdicts without the text either arm produced. The counts above were recorded at the time and are unaffected. Answers are now kept for 180 days.
Other measured skills in the registry, with their headline benchmark lift.