ClaudeCodeMod

All shelves / Mods / Workflow

actual-cli

By actual-software

actual-cli is a Claude Code workflow mod by actual-software.

For Actual AI, which keeps agents on a team's architecture decisions, draws a row of how many ADRs reached context and adds a file's rules beside each read.

Draws

Install

The marketplace name is actual-cli-skills.

In Claude Code
/plugin marketplace add actual-software/actual-skill
/plugin install actual-cli@actual-cli-skills
Open the repo

It runs inside the Claude Code process. The side panel is only what the author's README names. Read the repo before you install, or run claude plugin validate on the plugin folder.

Version
1.4.1
License
Apache-2.0
Repo stars
5
Last change
2026-10-16

What it can reach

  • Draws in the session
  • Reaches the network or an outside service
  • Sees tool calls
  • Calls another program on this machine

The row reads Actual AI: 3 of 12 ADRs added to context, shows 0 of N from the start, and updates after each reply in the background, so no reply waits on it. After /clear it counts the new session. It steps aside while a survey is showing, and [-] collapses the band. It shows nothing outside a repository with .actual/rules/, when the rules hold no decision, or when the CLI is missing or predates actual session summary. ACTUAL_PLAN_GATE=off leaves the row on. The same Claude Code as the briefing: 2.1.287 or later, or 2.1.285 and 2.1.286 with CLAUDE_CODE_ENABLE_FUNCTION_HOOKS set to 1. The band is in the terminal and the desktop app, not in VS Code or claude -p. Where the mod loads, briefing is capped at 48,000 characters per session by default, and the agent is told once when the cap is hit. A governed Read, or a read-only shell read, gets the rules for that file beside the result. A file written without being read first is briefed with the write. The model never reads the one-line transcript reports, and they are not counted against the cap. The skill runs actual adr-bot, which reads the repository and fetches the team's ADRs from Actual AI, then writes guidance into CLAUDE.md, AGENTS.md, or Cursor rules. An architecture question goes to the Actual advisor, which answers from those ADRs. Onboarding is at https://app.actual.ai, and the command reference is at https://actual.ai/cli/docs. The first /actual asks permission to npm install -g @actualai/actual. ACTUAL_RULES_BRIEF=off turns briefing off while the checks keep running. The author measured 212 local sessions: four in five file reads went through cat, head, sed -n, or grep rather than Read.

The marketplace name is actual-cli-skills. The marketplace source is ./. The row of ADRs added to context is why this page files it under workflow. jev-rules draws a pane of which project rules Jev gave Claude, and names them on a line under the prompt. claude-jev has Jev judge each edit against your instruction files and cite a broken rule by file and line, and sends a thin subagent brief back once. This one adds the team's rules for a file as Claude reads it, and counts the ADRs that reached context. The row and the briefing beside a read are this mod. The plan check, when the agent leaves plan mode, and the check of the turn's changes belong to the skill. The README says the transcript lines never touch that verdict. The actual command is the CLI the skill installs, a separate program.

Demo

The Session summary section of the Actual AI README: the row of ADRs added to context above the prompt, when it updates, and when it shows nothing.
Screenshot of the author's GitHub page.

What it hooks into

  • session.startWhen a session starts
  • turn.completeWhen a turn ends
  • classic.SessionStartA Claude Code event
  • ui.renderDraws into the sessionAbovePromptPane
  • command.runWhen you run its slash command/rules-brief
  • classic.PostToolUseA Claude Code event
  • classic.PreToolUseA Claude Code event
  • tool.callAround a tool call, where it can hold oneBashRead

Read from the on() calls in its hook modules. Open its source under Files to check what each one does.

Slash commands 1

  • /rules-briefWhat rule briefing has done this session, and where it stayed quiet

Hooks 6

  • SessionStart · startup|resume|clear|compact|fork"${CLAUDE_PLUGIN_ROOT}"/hooks/preflight.sh
  • UserPromptSubmit"${CLAUDE_PLUGIN_ROOT}"/hooks/turn-start.sh
  • PreToolUse · ExitPlanMode"${CLAUDE_PLUGIN_ROOT}"/hooks/plan-gate.sh
  • PreToolUse · Edit|Write"${CLAUDE_PLUGIN_ROOT}"/hooks/rules-brief.sh
  • PostToolUse · Read"${CLAUDE_PLUGIN_ROOT}"/hooks/rules-brief.sh
  • Stop"${CLAUDE_PLUGIN_ROOT}"/hooks/impl-gate.sh

Files

The author's README and source, 4 files. They load when you open them.

Facts

Author
actual-software
Category
Workflow
Marketplace
actual-cli-skills
Repo
actual-software/actual-skill
Version
1.4.1
License
Apache-2.0
Repo stars
5
Last change
2026-10-16
Language
Shell
Files
46

Other mods in Workflow

autodev-coredjnsty23Holds a commandFor Claude Auto-Dev, an autonomous workflow from brainstorm to ship, pins a status line under the prompt with what its hooks did and the sprint's pass states.kgai-modkgaidevDraws a status line when a decision is recorded, and Record and Skip in confirm mode, for kgai, the team's shared memory of why the code changed.glanceflowAntreas-StrbHolds a commandDraws a checklist above the prompt with the plan, the step Claude is on, how far along it is, and when Claude needs you.batonashishsk93Draws a band above the prompt of this session, the other sessions, and the tasks passed between them.todosbengousDraws the repository's TODOs in a scrolling band above the prompt, newest first, including findings an agent recorded.zcode-executorKyoMioFor zcode-executor, which hands Claude's tasks to the local ZCode agent, draws a live pane of this project's ZCode sessions and the requests waiting on you.arbeitslistenikisgeDraws a to-do field beside the chat, queues tasks while Claude works, and starts the next one, with Claude's plan above it.workfacescodge-24Draws the agent's notes in a panel and hands that file back verbatim after every compaction, auto or manual.brewmod-reminderskochetkov-maKeeps reminders as Markdown files that fire on time, draws a band of the fired ones in a later session, and has Claude run one when you pick it.navigatorqf-studioHolds a commandDraws a /nav pane of context, cost, reads, and the next task step as a button, and a phase band above the prompt, and guards fan-out reads.aftermidudevQueues prompts with /after while Claude works, draws the queue above the prompt, and sends the next one as your message when the turn ends.narratormidudevSpeaks a one or two sentence recap after each turn, in your language, with your system's best free voice, switched from the footer or /narrator.
All 396 in Workflow