Claude-Code v2.1.108 adds prompt caching controls, recap feature, and slash command discovery

New Environment Variables for Prompt Caching
v2.1.108 adds two environment variables for controlling prompt cache time-to-live (TTL). ENABLE_PROMPT_CACHING_1H opts into 1-hour prompt cache TTL on API key, Bedrock, Vertex, and Foundry deployments. The previously used ENABLE_PROMPT_CACHING_1H_BEDROCK is deprecated but still honored. FORCE_PROMPT_CACHING_5M forces a 5-minute TTL. The release also adds a warning at startup when prompt caching is disabled via DISABLE_PROMPT_CACHING* environment variables.
Session Recap Feature
A new recap feature provides context when returning to a session. This is configurable in /config and manually invocable with /recap. Users with telemetry disabled can force this feature with CLAUDE_CODE_ENABLE_AWAY_SUMMARY.
Skill Tool Enhancements
The model can now discover and invoke built-in slash commands like /init, /review, and /security-review via the Skill tool. This enables more dynamic interaction patterns where the AI can suggest relevant commands based on context.
Command Improvements
/undois now an alias for/rewind/modelnow warns before switching models mid-conversation, since the next response re-reads the full history uncached/resumepicker defaults to sessions from the current directory; pressCtrl+Ato show all projects
Error Message Improvements
Server rate limits are now distinguished from plan usage limits. 5xx/529 errors show a link to status.claude.com. Unknown slash commands suggest the closest match.
Performance and Memory Optimizations
Reduced memory footprint for file reads, edits, and syntax highlighting by loading language grammars on demand.
Interface Improvements
Added "verbose" indicator when viewing the detailed transcript (Ctrl+O). Fixed terminal escape codes appearing as garbage text in the prompt input after --teleport.
Bug Fixes
The release includes 16 bug fixes addressing issues ranging from paste functionality in /login prompts to session management problems:
- Fixed paste not working in the
/logincode prompt (regression in 2.1.105) - Fixed subscribers who set
DISABLE_TELEMETRYfalling back to 5-minute prompt cache TTL instead of 1 hour - Fixed Agent tool prompting for permission in auto mode when the safety classifier's transcript exceeded its context window
- Fixed Bash tool producing no output when
CLAUDE_ENV_FILE(e.g.~/.zprofile) ends with a#comment line - Fixed
claude --resume <session-id>losing the session's custom name and color set via/rename - Fixed session titles showing placeholder example text when the first message is a short greeting
- Fixed
/feedbackretry: pressing Enter to resubmit after a failure now works without first editing the description - Fixed
--teleportand--resume <id>precondition errors (e.g. dirty git tree, session not found) exiting silently instead of showing the error message - Fixed Remote Control session titles set in the web UI being overwritten by auto-generated titles after the third message
- Fixed
--resumetruncating sessions when the transcript contained a self-referencing message - Fixed transcript write failures (e.g., disk full) being silently dropped instead of being logged
- Fixed diacritical marks (accents, umlauts, cedillas) being dropped from responses when the
languagesetting is configured - Fixed policy-managed plugins never auto-updating when running from a different project than where they were first installed
📖 Read the full source: GitHub Claude-Code
👀 See Also

AI Agents Bet on World Cup: Why 'Keep Multiple Outcomes in Play' Wins
An experiment with 40+ AI agents placing real-money bets on Polymarket shows that profitable agents back more than one outcome per match. The difference: belief vs. action.

Georgia AI Data Center Drained 29M Gallons of Unmetered Water
QTS Fayetteville campus drew 29M gallons via two unauthorized water connections over 15 months, causing low pressure complaints. County waived fines, charged $147K retroactive.

Anthropic launches Claude Community Ambassadors program
Anthropic has launched the Claude Community Ambassadors program, which provides resources for organizing local developer meetups and connecting builders worldwide. The program is open to participants from any background and location.

Apple's AI Strategy and the Commoditization of Intelligence
The article argues that Apple's conservative approach to AI may be advantageous as intelligence becomes commoditized, with models like Gemma4 achieving 85.2% on MMLU Pro while running on phones, and OpenAI's Sora costing $15M daily against $2.1M revenue.