Skip to main content

codex-tracker-suite

This skill should be used to search, inspect, list, and recover local Codex sessions through Codex's native App Server APIs. Trigger for requests such as find a prior Codex conversation, search Codex transcript text, list recent Codex work, locate matches inside a session, or produce a safe Codex resume command. Do not use for Claude Code sessions, project source-code search, or web search.

Informações da origem

Repositório
tdimino/claude-code-minoan
Última atividade na origem
6 de setembro de 2026 às 20:58
Idioma detectado do SKILL.md
inglês
Estrelas
41
Forks
4

Opções de instalação

Por padrão, está selecionado o prompt que primeiro revisa a origem. Você pode mudar para um comando direto ou baixar uma cópia local.

Revise os arquivos de origem

Leia o SKILL.md e os arquivos complementares exibidos pelo SkillsMP antes de decidir se vai instalar.

Explorador de arquivos
7 arquivos

Exibindo SKILL.md

SKILL.md
Instruções da origem · Visualização somente leitura
name
codex-tracker-suite
description
This skill should be used to search, inspect, list, and recover local Codex sessions through Codex's native App Server APIs. Trigger for requests such as find a prior Codex conversation, search Codex transcript text, list recent Codex work, locate matches inside a session, or produce a safe Codex resume command. Do not use for Claude Code sessions, project source-code search, or web search.
# Codex Tracker Suite Use the bundled wrapper instead of reading Codex's private SQLite databases or rollout files. It talks to the version-matched `codex app-server` protocol and leaves Codex as the source of truth. ```bash codex-tracker search "topic" ``` ## Route the request - Find a session by topic or transcript text: `search QUERY` - Locate every visible match inside one session: `occurrences SESSION_ID QUERY` - Browse recent work: `recent` - Inspect metadata or the visible conversation: `show SESSION_ID` - See sessions loaded in a shared managed runtime: `alive` - Produce a shell-safe command for continuing work: `resume SESSION_ID` Run any command with `--help` for filters and JSON output. Add `--include-non-interactive` for `codex exec` and App Server sessions, or `--all-sources` for subagents too. Ordinary searches default to interactive CLI and IDE sessions, matching Codex's native behavior. ## Operating rules - Keep lookup operations read-only. `resume` prints a command; it does not launch an interactive process. - Return the matching session ID, title, working directory, snippet, and resume command when available. - Prefer `--json` when another program or agent will consume the result. - Native transcript search is experimental and may take a minute or two on a large history. The wrapper prints progress while waiting and uses a configurable timeout. - If native full-text search is unavailable, report the compatibility error and suggest `codex resume --all`; do not silently scrape private databases. - App Server may update its own state metadata during a read. In a restricted Codex sandbox, request approval for the wrapper command instead of bypassing the sandbox. - Do not archive, delete, rename, or otherwise mutate sessions unless the user separately asks for that operation. This wrapper intentionally exposes no destructive commands. - Treat `alive` as a managed-daemon view. Without the shared App Server daemon, use Codex's own session UI or process-level diagnostics instead. For protocol choices, compatibility behavior, and parity boundaries, read [references/architecture.md](references/architecture.md) only when maintaining or extending the wrapper.
Ver no GitHub