Skip to content
Back to skills

Convo

ASecurity

Recall past agent work, decisions, commands or session IDs across Claude Code and Codex transcripts with the indexed convo CLI, instead of searching transcript storage.

  • 3 stars
  • 0 votes
  • 0 copies
  • 1 view
  • Added October 13, 2026
ai-agentsbashsql

Works with

  • claude code
  • cli

Security analysis

A100/100

Pro scans all 3 files and shows the line behind each finding

Scanned October 13, 2026

npx -y skills add tompassarelli/north --skill convo --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Convo?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Convo
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/tompassarelli-convo/badge)](https://www.skillsdirectory.com/skills/tompassarelli-convo)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: convo
description: >-
  Recall past agent work, decisions, commands or session IDs across Claude Code and Codex transcripts with the indexed convo CLI, instead of searching transcript storage.
grounded: 2026-10-10
written: 2026-10-10
---

# Recall past work with convo

Never recursively scan transcript storage with raw tools; open what `convo` points at.

## Find
- `convo <words>` ranks sessions, each with its best message and an `[id]`.
  The words must share one message; when none does, sessions are ranked by
  how many words they cover.
- Tokens with punctuation (`c289-ko-bottom`, `tools/x.ts`) match as phrases.
  Use `-x '<literal>'` for exact strings, ids and error text.
- Narrow with `-r user|assistant|thinking|tool|system`, `-t Bash`, `-p <project>`,
  `-s <session or subagent id prefix>`, `--provider claude|codex` and
  `--since`/`--until 12h|3d|2w|YYYY-MM-DD`.
- `-r user` holds typed prompts and task briefs; injected harness text is `system`.
- Filters without words list the newest messages: `convo -r user --since 1d`.
- `-m` lists messages instead of sessions, `--json` feeds tools, `--full` prints
  whole messages.

## Open
- `convo show <id> -C 3` prints the message in full with its neighbours, session,
  path and line.
- `convo session <id prefix>` lists a session's transcripts, its first ask and the
  other sessions that mention it.
- Read a raw transcript only at the path and line that `show` printed.

## Judge
- Exit 0 means found, 1 a conclusive miss, and 2 inconclusive (index busy, root
  missing, `-u`) or bad usage. Retry or narrow after 2; it never proves absence.
- Only messages and tool calls are indexed, never tool output, so text seen only
  in a command's output is a miss.
- Every search refreshes the index first. Calls to convo itself are not results,
  and the current session ranks last.
- A hit is a recorded claim. Check the speaker, the date and any later reversal,
  and verify consequential facts against current source.
- For a verdict, add words like `landed`, `refuted` or `decided` to the topic.

Maintenance (`index --full`, `compress`, `restore`), the corpus layout and the
skill-usage SQL are in `references/notes.md`. Everything else is in `convo --help`.

Files in this skill

  • SKILL.md2.2 KB
  • agents/openai.yaml210 B
  • references/notes.md3.4 KB

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.