Cheatsheets
Quick reference commands for all ACFS tools. Click any command to copy.
Showing 7 of 7 tools
Git-native issue tracker. All state lives in .beads/ and is committed with code.
br ready --jsonShow unblocked open beads
br create "title" -t bug -p 1Create issue (type: bug/feature/task/epic/chore, priority: 0-4)
br update br-42 --claimClaim a bead (set in_progress + assignee)
br close br-42 --reason "Done"Close a bead with reason
br list -s open --jsonList beads by status
br show br-42 --jsonShow bead details
br update br-42 -p 0Set priority (0=critical, 4=backlog)
br sync --flush-onlyExport to .beads/issues.jsonl without git ops
- "ready" is a query, not a status -- it means open + unblocked by dependencies.
- Always commit .beads/ in the same commit as code changes.
- Priority 0 = critical (security, data loss). Default is 2.
Graph-aware triage engine for Beads. Computes PageRank, critical paths, cycles, and parallel tracks.
bv --robot-triageTHE mega-command: full triage with ranked recommendations
bv --robot-nextSingle top pick + claim command
bv --robot-planParallel execution tracks with unblocks lists
bv --robot-insightsFull graph metrics: PageRank, betweenness, HITS, cycles
bv --robot-alertsStale issues, blocking cascades, priority mismatches
bv --robot-suggestHygiene: duplicates, missing deps, label suggestions
bv --robot-triage --label backendScope triage to a label's subgraph
bv --robot-forecast allETA predictions with dependency-aware scheduling
- NEVER run bare `bv` -- it launches an interactive TUI that blocks your session.
- Always use --robot-* flags for deterministic JSON output.
- Phase 1 metrics are instant; Phase 2 (PageRank, etc.) has a 500ms timeout.
Procedural memory for agents. Extracts lessons from past sessions and retrieves them for new tasks.
cm context "task description" --jsonGet relevant rules, anti-patterns, and history for a task
cm playbook listList all playbook rules
cm playbook add "rule" --category "debugging"Add a new rule to the playbook
cm reflect --days 7Extract rules from recent sessions
cm doctor --fixHealth check and auto-repair
cm onboard statusCheck onboarding progress
cm onboard sample --fill-gapsGet sessions to analyze (filtered by playbook gaps)
- Run `cm context` before every non-trivial task.
- Context returns: relevantBullets, antiPatterns, historySnippets, suggestedCassQueries.
- Leave inline feedback: // [cass: helpful b-xyz] or // [cass: harmful b-xyz].
Indexes conversations from Claude Code, Codex, Cursor, Gemini, ChatGPT, etc. so solved problems can be reused.
cass search "query" --robot --limit 5Search past agent sessions
cass healthCheck indexing status
cass view /path -n 42 --jsonView specific message in a session
cass expand /path -n 42 -C 3 --jsonView message with 3 lines of context
cass capabilities --jsonList available search capabilities
cass robot-docs guideFull robot-mode documentation
- Never run bare `cass` -- it launches a TUI. Always use --robot or --json.
- stdout is data-only, stderr is diagnostics. Exit 0 = success.
- Use --fields minimal for lean output. Filter by agent with --agent.
Static analysis tool that flags likely bugs before commit. Exit 0 = safe, exit >0 = fix and re-run.
ubs file.tsScan specific file (<1 second)
ubs $(git diff --name-only)Scan all changed files
ubs $(git diff --name-only --cached)Scan staged files (pre-commit)
ubs .Full project scan (auto-ignores .venv, node_modules)
ubs --ci --fail-on-warning .CI mode -- strict, fails on warnings
ubs --only=js,python src/Language filter (3-5x faster)
- Golden rule: `ubs <changed-files>` before every commit.
- Scope to changed files for speed. `ubs src/file.ts` (<1s) vs `ubs .` (30s).
- Output format: file:line:col -- location, bulb icon = fix suggestion.
Agent cockpit for managing tmux sessions with multiple coding agents (Claude, Codex, Gemini).
ntm lsList all sessions
ntm new "name"Create a new session
ntm attach "name"Attach to an existing session
ntm send "name" "cmd"Send a command to a session
ntm kill "name"Kill a session
ntm healthCheck tool and daemon health
ntm quotaCheck agent quota usage
- Sessions persist across SSH disconnects.
- Use `ntm send` to dispatch prompts to agents without attaching.
- Health check verifies all dependencies (tmux, claude, codex, gemini CLIs).
Multi-repo sync tool with AI-driven commit automation. Manages cloning, pulling, and automated commits across all repositories.
ru syncClone missing + pull updates for all repos
ru sync --parallel 4Parallel sync with 4 workers
ru status --fetchFetch + show ahead/behind for each repo
ru agent-sweep --dry-runPreview dirty repos to process
ru agent-sweep --parallel 4AI-driven commits in parallel
ru list --pathsList all managed repo paths
- Exit codes: 0=success, 1=partial failure, 2=conflicts, 5=interrupted (use --resume).
- Always `ru status` before `ru sync` to preview changes.
- Use --repos=pattern to scope operations to matching repos.