From jacked
Use when the user mentions a past project like "configurator", asks to continue/resume previous work, says "how did I do X before", references past sessions, or starts work on a feature that may have been done before. Searches and loads context from past Claude Code sessions.
npx claudepluginhub jackneil/claude-jacked --plugin jackedThis skill uses the workspace's default tool permissions.
Search and load context from past Claude Code sessions using semantic search.
Guides strict Test-Driven Development (TDD): write failing tests first for features, bugfixes, refactors before any production code. Enforces red-green-refactor cycle.
Guides systematic root cause investigation for bugs, test failures, unexpected behavior, performance issues, and build failures before proposing fixes.
Guides A/B test setup with mandatory gates for hypothesis validation, metrics definition, sample size calculation, and execution readiness checks.
Search and load context from past Claude Code sessions using semantic search.
This skill requires the search extra (jacked[search]) and a configured Qdrant instance. Before doing anything else, check if search is available:
jacked search "test" --limit 1
If this returns an error like 'search' requires the search extra or fails to connect to Qdrant, STOP and tell the user:
"The /jacked session search feature isn't set up yet. It requires
uv tool install "claude-jacked[search]"and a Qdrant instance. For now, I can check git history or local session files instead."
Do NOT attempt to install the search extra automatically.
/jacked <description of what you want to work on>
Example:
/jacked implement overnight OB time handling
When the user runs /jacked <description>, follow these steps:
Run the search command:
jacked search "<user's description>" --limit 5
The output includes:
Use the AskUserQuestion tool with multiSelect=true to let the user pick which sessions to load:
{
"questions": [{
"question": "Which sessions would you like to load context from?",
"header": "Sessions",
"multiSelect": true,
"options": [
{
"label": "1. YOU - 24d ago ๐๐ค",
"description": "hank-coder: Implementing overnight time calculation..."
},
{
"label": "2. @bob - 3mo ago ๐ค",
"description": "hank-coder: Time handling refactor for multiple..."
},
{
"label": "3. YOU - 2d ago",
"description": "krac-llm: Staff time merging edge cases..."
},
{
"label": "None - skip",
"description": "Don't load any previous context"
}
]
}]
}
Note: AskUserQuestion supports max 4 options, so if there are 5+ results, show top 3 + "None" option.
When the user selects one or more sessions:
# Default: smart mode (recommended) - plan + summaries + labels + user msgs
jacked retrieve <session_id> --mode smart
# For full transcript (use sparingly - can be 50K+ chars)
jacked retrieve <session_id> --mode full
Smart mode retrieval output includes:
If session is local: Tell the user:
This session exists locally on your machine!
To resume it natively (with full Claude memory), run in a new terminal:
claude --resume <session_id>
Or I can inject the smart context into our current conversation.
Would you like me to inject it here? (yes/no)
If session is remote only: Tell the user:
This session is from another machine (<machine_name>).
I'll inject the context into our conversation.
The retrieve output already includes proper formatting with staleness warnings.
For context older than 7 days, include the staleness warning that appears in the output:
After injection, summarize:
| Mode | What's Included | When to Use |
|---|---|---|
| smart | Plan + agent summaries + labels + user msgs | Default - best balance |
| plan | Just the plan file | Quick strategic overview |
| labels | Just summary labels (tiny) | Quick topic check |
| agents | All subagent summaries | Deep dive into exploration results |
| full | Everything including transcript | Need full details (use sparingly) |
jacked backfill to index your sessions."uv tool install claude-jacked to install."claude --resume for the best experiencejacked cleardb to wipe your data before re-indexing with a new schema