By Enki-IQ
Codebrief intent capture — distills developer intent locally at session end, after a git push, or after a PR is created, and sends only derived annotations (never code) to your Codebrief workspace.
Turn off Codebrief intent capture for the current repository.
Turn on Codebrief intent capture for the current repository.
Authenticate the Codebrief capture CLI by authorizing it in your browser. Use when capture says you're not logged in.
Clear the Codebrief capture CLI credentials and stop capturing.
Show whether the Codebrief capture CLI is logged in.
Distills developer decisions, plans, deferrals, and constraints from Claude Code or Codex sessions, then sends only bounded and scrubbed intent annotations to your Codebrief workspace. It never sends source code or transcripts to Codebrief.
codex plugin marketplace add Enki-IQ/codebrief-capture
codex plugin add codebrief-capture@codebrief
Start a new Codex thread after installation so the skills and hooks are loaded. The plugin works in Codex CLI and in Conductor Codex workspaces without extra Conductor configuration.
claude plugin marketplace add Enki-IQ/codebrief-capture
claude plugin install codebrief-capture@codebrief
It works in any terminal that runs Claude Code, including integrated IDE terminals.
codebrief-login skill. Browser authorization is preferred;
the fallback hidden prompt keeps a pasted key out of the conversation and shell history.codebrief-enable or codebrief-disable skill to change the current repository.codebrief-status to check login, host CLI availability, distillation model, and
enabled repositories. Use codebrief-logout to remove local credentials.Claude captures at session end and in the background after a successful git push or
gh pr create. Codex captures after 45 seconds of end-of-turn inactivity and immediately
after those same successful publish commands. Repeated Stop events and publish hooks for
the same transcript revision are deduplicated.
Claude login/enable also adds a Codebrief: capturing (or off) segment to Claude
Code's status line. The Codex package does not modify Claude settings.
No transcript content is sent to Codebrief. Distillation context goes only to the
user's configured host provider. The Codex package structurally reduces the local
rollout first, excluding system instructions, reasoning, tool output, images, and world
state, then sends that reduced text through an ephemeral, read-only codex exec child
with hooks and plugins disabled and MCP servers cleared. Claude sends a bounded recent
transcript tail through an equivalent non-persistent claude -p child under the user's
Anthropic account.
The derived records are scrubbed and bounded again before Codebrief ingestion. Provider calls run under the user's own Claude or Codex account; Codebrief is never billed for distillation.
Claude defaults to sonnet; override distillModel in ~/.codebrief/config.json or
set CODEBRIEF_DISTILL_MODEL. Codex uses the user's default model with low reasoning;
override codexDistillModel, codexDistillReasoningEffort, or the corresponding
CODEBRIEF_CODEX_DISTILL_MODEL and CODEBRIEF_CODEX_REASONING_EFFORT environment
variables. codexDebounceMs or CODEBRIEF_CODEX_DEBOUNCE_MS controls the idle delay.
Executes bash commands
Hook triggers when Bash tool is used
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
npx claudepluginhub enki-iq/codebrief-capture --plugin codebrief-captureLazy senior dev mode. Forces the simplest, shortest solution that actually works: YAGNI, stdlib first, no unrequested abstractions.
Consult multiple AI coding agents (Gemini, OpenAI, Grok, Perplexity, plus codex, antigravity, and grok CLIs when installed) to get diverse perspectives on coding problems
Production-grade engineering skills for AI coding agents — covering the full software development lifecycle from spec to ship.
Evidence-gated AI coding workflow: scan → analyze → plan → TDD → execute → fix → verify → review, powered by Codebase Memory MCP >= 0.9.0 with optional Serena LSP intelligence. Includes blast-radius planning, test/cycle gates, independent review, and Windows Git Bash hook auto-resolution.
Feature development with code-architect/explorer/reviewer agents, CLAUDE.md audit and session learnings, and Agent Skills creation with eval benchmarking from Anthropic.
Permanent coding companion for Claude Code — survives any update. MCP-based terminal pet with ASCII art, stats, reactions, and personality.