Unnamed Skill

Search past conversations by keywords, timeframe, or ID. Triggers on "search sessions", "find that conversation", "what did we work on", "look up past discussion about X", "where were we", "what did I learn", "knowledge gaps", "retrospective". Also trigger on: "what did we discuss...", "you mentioned...", past tense verbs referring to prior work, possessives without context ("my project", "my auth system"), and assumptive questions that reference unshared context. Extracts clean conversation data and applies analytical lenses for structured insights.

$ Installer

git clone https://github.com/gupsammy/Claude-setup /tmp/Claude-setup && cp -r /tmp/Claude-setup/skills/session-search ~/.claude/skills/Claude-setup

// tip: Run this command in your terminal to install the skill


name: session-search description: Search past conversations by keywords, timeframe, or ID. Triggers on "search sessions", "find that conversation", "what did we work on", "look up past discussion about X", "where were we", "what did I learn", "knowledge gaps", "retrospective". Also trigger on: "what did we discuss...", "you mentioned...", past tense verbs referring to prior work, possessives without context ("my project", "my auth system"), and assumptive questions that reference unshared context. Extracts clean conversation data and applies analytical lenses for structured insights.

session-search

Extraction

python3 ~/.claude/skills/session-search/scripts/extract_conversations.py --days 3
OptionEffect
--days NDays from last activity (not today)
--from-todayDays from today instead
--all-projectsCross-project (implies --from-today)
--project /pathFilter to specific project
--compactNo metadata, more conversation
--min-exchanges NSkip sessions with < N exchanges
--ids abc,defFetch specific conversations
--paths /path.jsonlDirect file paths

Output: Default shows files, tools, errors + conversation. --compact omits metadata.

Time modes: --days N counts from last activity (useful when returning to old projects). --all-projects or --from-today counts from today (calendar-based).

Workflow

  1. Grep for keywords (generate semantic synonyms):

    grep -l -E "auth|login|session" ~/.claude/projects/*/*.jsonl
    
  2. Extract: python3 ... --paths /found/conv.jsonl

  3. Apply lens using parameters and questions below.


Lenses

Routing

User SaysLens
"where were we", "recap"restore-context
"what I learned", "reflect"extract-learnings
"gaps", "struggling"find-gaps
"mentor", "review process"review-process
"retro", "project review"run-retro
"decisions", "CLAUDE.md"extract-decisions
"bad habits", "antipatterns"find-antipatterns

Parameters

LensDaysFlagsAlso Gather
restore-context3git status, git log -10
extract-learnings14--all-projects --compact
find-gaps30--all-projects --compact
review-process14--all-projects --compactrecent git log
run-retro30--project /pathfull git history
extract-decisions90--project /path
find-antipatterns30--all-projects --compact

--min-exchanges 2 or 3 filters out short sessions and reduces noise.

Core Questions

LensAsk
restore-contextWhat's unfinished? What were the next steps?
extract-learningsWhere did understanding shift? What mistakes became lessons?
find-gapsWhat topics recur? Where is guidance needed repeatedly?
review-processIs there planning before coding? Is debugging systematic?
run-retroHow did the solution evolve? What worked? What was painful?
extract-decisionsWhat trade-offs were discussed? What was rejected and why?
find-antipatternsWhat mistakes repeat? What confusions persist?

Follow-ups: find-gaps → suggest learn-anything. extract-decisions → suggest /updateclaudemd.

Grep Signals

Use these patterns to find relevant sessions before extracting:

LensGrep Pattern
extract-learningslearned|realized|understand now|clicked|got it
find-gapsconfused|don't understand|struggling|help with
extract-decisionsdecided|chose|instead of|trade-off|because
find-antipatternsagain|same mistake|repeated|forgot

Synthesis

Principles

  1. Prioritize significance — 3-5 key findings, not exhaustive lists
  2. Be specific — file paths, dates, project names
  3. Make it actionable — every finding suggests a response
  4. Show evidence — quotes or references
  5. Keep it scannable — clear structure, no walls of text

Structure

## [Analysis Type]: [Scope]

### Summary
[2-3 sentences]

### Findings
[Organized by whatever fits: categories, timeline, severity]

### Patterns
[Cross-cutting observations]

### Recommendations
[Actionable next steps]

Length

Default: 300-500 words. Expand only when data warrants it.