Use when adding new error messages to React, or seeing "unknown error code" warnings.
npx skills add antoniolg/agent-kit --skill "total-recall"
Install specific skill from multi-skill repository
# Description
Search past conversations and sessions with Antonio. Use when he asks to remember/find something discussed before, what was decided about X, or to look back at previous sessions. Triggers: 'recuerdas cuando...', '¿qué decidimos sobre...?', 'busca en conversaciones anteriores', 'find where we discussed...'. Requires contextify CLI.
# SKILL.md
name: total-recall
description: "Search past conversations and sessions with Antonio. Use when he asks to remember/find something discussed before, what was decided about X, or to look back at previous sessions. Triggers: 'recuerdas cuando...', '¿qué decidimos sobre...?', 'busca en conversaciones anteriores', 'find where we discussed...'. Requires contextify CLI."
Contextify Total Recall
Search your conversation history to find past decisions, solutions, and discussions.
Output Format
Begin your response with:
Contextify Total Recall
Then provide the search results with citations.
Trigger phrases
- "use total recall to..."
- "search our conversation history"
- "find where we discussed..."
- "what did we decide about..."
- "look back in past sessions"
- "remember when we talked about..."
Preconditions
This skill requires the Contextify app and CLI.
It works with Claude Code and Codex CLI.
1) Check CLI availability:
command -v contextify
2) If contextify is missing:
Error response:
Contextify CLI not found.
DMG users: Open Contextify → Settings → CLI → "Install/Repair CLI"
App Store users: Run
brew install PeterPym/contextify/contextify-queryFor help: https://contextify.sh/help
Canonical loop
1) Confirm database availability:
contextify status --json
If database not found, respond:
Contextify database not found.
Please open Contextify once to initialize the database.
Download: https://contextify.sh/download
2) Search for an anchor:
contextify search "<query>" --project . --days 30 --limit 10 --json
Anchor selection guidance:
- If asking about earlier context (not "in this chat"), prefer anchors NOT from the active transcript.
- If
CONTEXTIFY_CLAUDE_TRANSCRIPT_IDis set, treat hits from that transcript as lower priority unless user explicitly wants current session. - If transcript ID missing and multiple hits exist, avoid auto-selecting anchors from last 30 minutes.
3) Retrieve context around the anchor:
contextify context "<entry-uuid>" --before 10 --after 20 --project . --json
4) Format response:
Contextify Total Recall
Found: [brief summary of what was found]
From: [date/time and project context]
[Key excerpts with citations]
Entry ID:
<uuid>(for reference)
Error handling
| Error | Response |
|---|---|
dbNotFound |
"Contextify database not found. Open Contextify to initialize. https://contextify.sh/download" |
dbProjectNotFound |
Check details.suggestions, offer alternatives |
featureUnavailable |
Explain limitation clearly, do not imply workarounds |
entryNotFound |
Re-search for a new anchor |
cliNotFound |
"Contextify CLI not found. See https://contextify.sh/help for installation." |
Expanding search
If search returns 0 results:
1. Widen --days (try 90 or 365)
2. If not clearly about current repo, retry without --project .
3. Use contextify projects --json to discover other projects
4. Ask user to clarify what they're looking for
Delegating to researcher agent
Claude Code only: For complex multi-query searches, delegate to contextify-researcher agent:
Use the contextify-researcher agent to thoroughly search for [topic]
The researcher agent will perform multiple searches and synthesize results.
Codex CLI: Agent delegation is not available. Instead, run multiple searches manually and synthesize results yourself.
# Supported AI Coding Agents
This skill is compatible with the SKILL.md standard and works with all major AI coding agents:
Learn more about the SKILL.md standard and how to use these skills with your preferred AI coding agent.