Recall and summarize recent Claude Code sessions for the current working directory. Parses JSONL session logs to bring previous conversation context into the current session. Use when the user says "/recall", "/review", "/recap", "what did we do last time", "what was I working on", "recall previous session", "review recent work", or wants context from prior Claude sessions.
How this skill is triggered — by the user, by Claude, or both
Slash command
/claude-session-recall:session-recallThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Load and summarize recent Claude Code sessions for the current working directory (including subdirectories). This gives you context about what was discussed, what files were changed, and what commands were run in prior sessions.
Load and summarize recent Claude Code sessions for the current working directory (including subdirectories). This gives you context about what was discussed, what files were changed, and what commands were run in prior sessions.
python3 ${CLAUDE_SKILL_DIR}/scripts/recall_cli.py --cwd "$(pwd)" ```
If the user specifies a different time range (e.g., "last week"), adjust `--days` accordingly:
```bash
python3 ${CLAUDE_SKILL_DIR}/scripts/recall_cli.py --cwd "$(pwd)" --days 7 ```
If the user wants more sessions, adjust `--min-sessions`:
```bash
python3 ${CLAUDE_SKILL_DIR}/scripts/recall_cli.py --cwd "$(pwd)" --min-sessions 5 ```
To adjust the token budget (default 40k):
```bash
python3 ${CLAUDE_SKILL_DIR}/scripts/recall_cli.py --cwd "$(pwd)" --max-tokens 80000 ```
### Step 2: Summarize
Read the output and provide a **concise summary** to the user:
1. **Timeline**: When each session happened and how long it lasted
2. **What was done**: The key tasks, questions asked, and outcomes in each session
3. **Files changed**: Any files that were created or modified
4. **Open threads**: Any unfinished work, unanswered questions, or next steps that were mentioned
Keep the summary brief and actionable. Focus on what's most relevant for continuing work.
## Notes
- The script searches `~/.claude/projects/` for session logs matching the current working directory
- It includes sessions from subdirectories of the current cwd
- Selection logic: max(last 2 days, last 2 sessions) — configurable via `--days` and `--min-sessions`
- The current (most recent) session is excluded by default; use `--include-current` to override
- Subagent logs are merged chronologically into their parent session if main sessions fit under the token budget
- Output is capped at 40k tokens (configurable via `--max-tokens`); oldest messages are trimmed first, then entire sessions are dropped if needed
- Uses tiktoken for approximate token counting (requires `pip3 install tiktoken`)
- Progress/debug output goes to stderr; structured output goes to stdout
npx claudepluginhub bw2/claude-session-recallExtracts user inputs from Claude Code session history and organizes them into date-based markdown files in a project's .chats directory for documentation and reference.
Analyzes Claude Code session history JSONL files to extract insights, summaries, and patterns from conversations. Processes current project or all sessions with bash, jq, and subagents.
Searches and recovers content from Claude Code JSONL session history across active config homes and registered archives. Useful for keyword/date-bounded history search, prior-conversation forensics, deleted-file recovery, and tool/file-operation analysis.