Execute development tasks using OpenAI Codex CLI for code generation, refactoring, feature implementation, and bug fixes.
Use ATM as the out-of-session substrate: spawn Claude/Codex panes running /rpi and /evolve over a bead queue, then tend the swarm to convergence.
Use when debugging native programs or ELF binaries with gdb breakpoints, backtraces, and memory inspection. Triggers:
Use when using ru for multi-repo commits, sync, GitHub review, or maintenance automation.
Get a help or a second opinion from other agents. Bundle a prompt and a curated file set, then ask other LLMs for advice (debug, refactor, design, ...).
Research software tools via source code, GitHub, web. Use when creating skills, learning new tools, finding undocumented features, or bleeding-edge patterns.
Use when rationalizing file or directory layout and updating references without breaking builds. Triggers:
Improve OpenShift/Kubernetes Grafana dashboards and validate PromQL/GitOps changes.
Use when operating an NTM review or analysis worker with bounded inputs and evidence-backed output. Triggers:
Use when designing or auditing CLI doctor commands, health checks, repair hints, and diagnostic UX. Triggers:
Mine patterns that recur across multiple projects and generalize into reusable artifacts. Use when "I've seen this before", DRY across repos, or building shared libraries.
Repeatable Opus (Claude) + Codex dual-pane ATM collaboration. Triggers: "dual pane", "Opus and Codex together", "CEP duel/build", "two-pane ATM", "collaborative ATM".
Use when auditing Rust unsafe blocks and FFI boundaries, invariants, tests, and tooling. Triggers:
Use when auditing git stashes, deciding keep/drop/apply/archive, and clearing confirmed stale entries. Triggers:
Use when Codex needs Fable approval through an ATM/NTM validator pane. Triggers: - codex approval - ask fable - fable plan review
Compile or lint a persistent Mayor-style goal prompt that ratchets a bead graph through bounded RPI experiments toward one larger outcome.
Use when coordinating browser or UI tests through NTM panes with screenshots and handoffs. Triggers:
Use when operating ACFS flywheel health checks, init, and agent loop tooling from ~/acfs/bin/acfs. Triggers:
Use when improving CLI ergonomics for agents: flags, help, JSON output, exit codes, and robot surfaces. Triggers:
Use when restoring machine responsiveness from high CPU, memory, IO, cache, or runaway process pressure.
Use when implementing a written spec into a reliable service with acceptance examples and observability. Triggers:
Measure declared project fitness goals without recommending or applying work. Triggers: "fitness", "check project fitness", "measure goals".
Mid-epic strategic drift audit: code is ground truth, README/PRODUCT.md/plan docs are the measuring stick; emit a cited gap report and a routing decision.
Drive AgentOps in AGY: loop, plugins, memory, evidence, scoped worktrees. Triggers: agy, antigravity, agy plugin, AGY evidence.
Fully autonomous epic execution. Runs until ALL children are CLOSED. Loops through beads issues, runs /implement on each, validates with /vibe. NO human prompts, NO stopping.
Use when isolating parallel Claude Code workers in separate git worktrees to prevent file collisions. Triggers:
Use when running codex exec in a least-privilege sandbox with machine-checkable proof. Triggers:
Use when designing fuzz, property, randomized, or corpus-based tests and replaying failures. Triggers:
Restore system responsiveness via safe, ordered process cleanup and agent-swarm hygiene.
Grade agent or model output against Outcomes for holdout-safe evals and runtime comparisons.
Test repeated implementation shapes against independent exemplars and a holdout before routing an earned abstraction.
Wire MCP servers and AgentOps plugin bundles into the AGY image with least-privilege access, rollback evidence, and validation hooks.
Use when cleaning repository branches, worktrees, gc state, large objects, and exclusion rules safely. Triggers:
Run AGY headlessly via scheduled ticks or `agy -p`, capture agentapi JSONL evidence, and validate automated AGY loops or event streams.
Research an upstream repo before contributing: rules, patterns, and expectations.
Operate the Agentic Coding Flywheel as a caller-selected software factory; keep its runtime state out of AgentOps verdicts. Triggers: "using flywheel", "agent flywheel".
Use when wiring MCP servers or plugins into Codex CLI and the AgentOps Codex skill bundle. Triggers:
Use when auditing codebase risks with evidence and prioritized remediation. Triggers:
Use when producing a shareable architecture, module, metrics, and health report for a codebase. Triggers:
Investigate unfamiliar legacy code before edits. Triggers: legacy module, unknown repo, risky refactor, trace ownership. Triggers:
Use when dispatching scoped Claude Code subagents with worktrees, roles, tools, memory, and evidence gates. Triggers:
Configure Claude Code hooks for PreToolUse, PostToolUse, Stop, Notification. Use when blocking commands, auto-formatting, custom permissions, or writing hooks.
Coding Agent Session Search - unified CLI/TUI to index and search local coding agent history from Claude Code, Codex, Gemini, Cursor, Aider, ChatGPT, Pi-Agent, Factory, and more.
Graph-aware task triage with bv and br. Use when prioritizing work, finding bottlenecks, tracking dependencies, or managing local issues across projects.
Consolidate fragmented agent-memory layers into one bd-canonical store, then GC/retire the rest. Triggers: "memory migration", "consolidate agent memory", "beads-first memory".
Install AGY rules, workflow, goal, and schedule controls for AgentOps loop law. Triggers: AGY rules, agy-loop, AGY schedule.
Use when designing Rust CLIs backed by SQLite with migrations, transactions, tests, and data safety. Triggers:
Cross-validate decisions using multiple AI models (Codex, Gemini, Grok). Use when "get a second opinion", evaluating approaches, or high-stakes decisions.
Use when turning tacit expert know-how into a durable skill, playbook, or checklist. Triggers:
Convert AgentOps skill formats. Triggers: "converter", "convert agentops skill formats.", "converter skill".
Own the unattended renewal spine: renewal ticks, the two-tick stall rule, and escalation for NTM panes coordinated over MCP Agent Mail.
Use when adding fast lexical or semantic code search to a Rust project with an ergonomic query API. Triggers:
Use when removing generic filler from code, docs, or handoffs while preserving every load-bearing fact. Triggers:
Use when simplifying code, reducing duplication, or clarifying flow while preserving behavior with tests. Triggers:
Use when using Codex Goals to define an objective once and let Codex iterate until done. Triggers:
Use when finding and fixing deadlocks with lock ordering, reproduction, timeouts, or lock-free alternatives. Triggers:
Local-first issue tracker (beads_rust) for AI agents. Use when tracking tasks, managing dependencies, finding ready work, or syncing issues to git via JSONL.
Use when searching code with rg using precise, fast flags instead of slow grep or find patterns. Triggers:
Use when finding mocks, stubs, fake paths, or placeholders leaking into production code. Triggers:
Prove AGY project/worktree isolation with scoped --add-dir permissions, role tiers, dcg guardrails, and persisted evidence.