Index of every page on this site: llms.txt
All skills
obra avatar

/remembering-conversations

@2db8793
by Jesse Vincentobra/clank43 stars
26
Run it yourself

Search previous Claude Code conversations for facts, patterns, decisions, and context using semantic or text search

Use this Skill: https://skilld.dev/gh/obra/clank/remembering-conversations

Nothing lands on disk. Nothing to clean up.

Fork this Skill

Edit a local copy. It keeps the author and licence.

SKILL.md

≈36 tokens for metadata: the name and description. ≈415 when used: this file. ≈29k more on demand in 6 files.

Description uses 5.0% of example budget

Before choosing a Skill, your Agent reads its name and description. All available Skills share that space.

  • A shorter description leaves more room for other Skills. This entry exceeds our 1% size suggestion.

In our Claude Code example, all Skill names and descriptions share 8,000 characters. This Skill uses ≈398 characters, or 5.0%.

The 1% threshold is a size suggestion. Longer descriptions can still fit.

Your model, settings, and other Skills decide how much text your Agent can read.

Example settings and source

The example uses a 200k-token context and default Claude Code settings. The count includes the name, description, separators, and when_to_use when present. Codex also counts local file paths.

Skit's source and limits: Codex 0.160.1, Claude Code 2.1.292.

Remembering Conversations

Search archived conversations using semantic similarity or exact text matching.

Core principle: Search before reinventing.

Announce: "I'm searching previous conversations for [topic]."

Setup: See INDEXING.md

When to Use

Search when:

  • Your human partner mentions "we discussed this before"
  • Debugging similar issues
  • Looking for architectural decisions or patterns
  • Before implementing something familiar

Don't search when:

  • Info in current conversation
  • Question about current codebase (use Grep/Read)

In-Session Use

Always use subagents (50-100x context savings). See skills/getting-started for workflow.

Manual/CLI use: Direct search (below) for humans outside Claude Code sessions.

Direct Search (Manual/CLI)

Tool: ~/.claude/skills/collaboration/remembering-conversations/tool/search-conversations

Modes:

search-conversations "query"              # Vector similarity (default)
search-conversations --text "exact"       # Exact string match
search-conversations --both "query"       # Both modes

Flags:

--after YYYY-MM-DD    # Filter by date
--before YYYY-MM-DD   # Filter by date
--limit N             # Max results (default: 10)
--help                # Full usage

Examples:

# Semantic search
search-conversations "React Router authentication errors"

# Find git SHA
search-conversations --text "a1b2c3d4"

# Time range
search-conversations --after 2025-09-01 "refactoring"

Returns: project, date, conversation summary, matched exchange, similarity %, file path.

For details: Run search-conversations --help

Source: SKILL.md on GitHub

skilld matched fixed text patterns in SKILL.md and file names. Patterns miss obfuscated code.

skilld run checks every file with the same patterns. It asks for approval before it loads a Skill with a behavior marked Needs approval.

No alerts22d3 checks · Risk SAFE
  • Gen Agent Trust Hub22d

    This skill provides a mechanism to index and search past conversation histories using vector similarity and AI-generated summaries. It includes tools for manual search, automatic indexing via session-end hooks, and a subagent template for synthesizing findings. All identified behaviors, including the use of hooks for automation and the Anthropic API for summarization, are consistent with the skill's primary documented purpose.

  • Socket22d

    No alerts

  • Snyk22d

    Risk: LOW · No issues

Signed by skilld at 2db8793. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub last month.

Dormantupdated last year
version
1.0.0
Other metadata
when_to_use
When your human partner mentions "we discussed this before". When debugging similar issues. When looking for architectural decisions or code patterns from past work. Before reinventing solutions. When you need to find a specific git SHA or error message.
  • Debugging
  • conversation-search
  • semantic-search
  • context-retrieval
  • claude-code
  • decision-tracking
  • architecture

README badge

README badge for obra/clank

Searches archived Claude Code conversations by semantic similarity or exact text match to recover facts, architectural decisions, code patterns, and debugging context from past sessions. Use it when your human mentions "we discussed this before" or when debugging recurring issues to avoid reinventing solutions.

Generated from the current SKILL.md.

How do I search conversations outside of Claude Code?
Use the CLI tool at ~/.claude/skills/collaboration/remembering-conversations/tool/search-conversations with a query string. Supports vector similarity (default), exact text match (--text flag), or both modes.
What's the difference between semantic and text search?
Semantic search finds conceptually similar conversations; text search matches exact strings like git SHAs or error messages. Use --text for precise matches and the default mode for topic-based lookups.
Can I filter search results by date?
Yes. Use --after YYYY-MM-DD and --before YYYY-MM-DD flags to narrow results to a specific time range.
When should I use this skill instead of searching the current conversation or codebase?
Search previous conversations when looking for architectural decisions, past code patterns, or context from earlier sessions. Use Grep/Read for info already in the current conversation or codebase.
Does this skill require setup?
Yes. Refer to INDEXING.md in the skill directory for initial setup instructions before using the search tool.

Generated from the current SKILL.md. These answers refresh after source changes.