shared

$27

All skills that reference external CLIs MUST degrade gracefully when those CLIs are absent.

Check Pattern

# Before using any external CLI, check availability

if command -v bd &>/dev/null; then

  # Full behavior with bd

else

  echo "Note: bd CLI not installed. Using plain text tracking."

  # Fallback: use TaskList, plain markdown, or skip

fi

Fallback Table

Capability

When Missing

Fallback Behavior

bd

Issue tracking unavailable

Use TaskList for tracking. Note "install bd for persistent issue tracking"

ao

Knowledge flywheel unavailable

Write learnings to .agents/learnings/ directly. Skip flywheel metrics

gt

Workspace management unavailable

Work in current directory. Skip convoy/sling operations

codex

CLI missing or model unavailable

Fall back to runtime-native agents. Council pre-flight checks CLI presence (which codex) and model availability for --mixed mode.

cass

Session search unavailable

Skip transcript search. Note "install cass for session history"

Model tier config

.agentops/config.yaml missing

Use built-in defaults (quality=opus, balanced=sonnet, budget=haiku). Tier resolution falls through to "balanced".

Required Multi-Agent Capabilities

Council, swarm, and crank require a runtime that provides these capabilities. If a capability is missing, the corresponding feature degrades.

Capability

What it does

If missing

Spawn subagent

Create a parallel agent with a prompt

Cannot run multi-agent. Fall back to --quick (inline single-agent).

Agent-to-agent messaging

Send a message to a specific agent

No debate R2. Workers run fire-and-forget.

Broadcast

Message all agents at once

Per-agent messaging fallback.

Graceful shutdown

Request an agent to terminate

Agents terminate on their own when done.

Shared task list

Agents see shared work state

Lead tracks manually.

Every runtime maps these capabilities to its own API. Skills describe WHAT to do, not WHICH tool to call.

After detecting your backend (see Backend Detection below), load the matching reference for concrete tool call examples:

Backend

Reference

Claude feature contract

skills/shared/references/claude-code-latest-features.md

Claude Native Teams

skills/shared/references/backend-claude-teams.md

Codex Sub-Agents / CLI

skills/shared/references/backend-codex-subagents.md

Background Tasks (fallback)

skills/shared/references/backend-background-tasks.md

Inline (no spawn)

skills/shared/references/backend-inline.md

Backend Detection

Use capability detection at runtime, not hardcoded tool names. The same skill must work across any agent harness that provides multi-agent primitives. If no multi-agent capability is detected, degrade to single-agent inline mode (--quick).

Selection policy (runtime-native first):

• If running in a Claude session and TeamCreate/SendMessage are available, use Claude Native Teams as the primary backend.

• If running in a Codex session and spawn_agent is available, use Codex sub-agents as the primary backend.

• If both are technically available, pick the backend native to the current runtime unless the user explicitly requests mixed/cross-vendor execution.

• Only use background tasks when neither native backend is available.

Operation

Codex Sub-Agents

Claude Native Teams

OpenCode Subagents

Inline Fallback

Spawn

spawn_agent(message=...)

TeamCreate + Task(team_name=...)

task(subagent_type="general", prompt=...)

Execute inline

Spawn (read-only)

spawn_agent(message=...)

Task(subagent_type="Explore")

task(subagent_type="explore", prompt=...)

Execute inline

Wait

wait(ids=[...])

Completion via SendMessage

Task returns result directly

N/A

Retry/follow-up

send_input(id=..., message=...)

SendMessage(type="message", ...)

task(task_id="<prior>", prompt=...)

N/A

Cleanup

close_agent(id=...)

shutdown_request + TeamDelete()

None (sub-sessions auto-terminate)

N/A

Inter-agent messaging

send_input

SendMessage

Not available

N/A

Debate (R2)

Supported

Supported

Not supported (no messaging)

N/A

OpenCode limitations:

• No inter-agent messaging — workers run as independent sub-sessions

• No debate mode (--debate) — requires messaging between judges

• --quick (inline) mode works identically across all backends

Backend Capabilities Matrix

Prefer native teams over background tasks. Native teams provide messaging, redirect, and graceful shutdown. Background tasks are fire-and-forget with no steering — only a speedometer and emergency brake.

Capability

Codex Sub-Agents

Claude Native Teams

Background Tasks

Observe output

wait() result

SendMessage delivery

TaskOutput (tail)

Send message mid-flight

send_input

SendMessage

NO

Pause / resume

NO

Idle → wake via SendMessage

NO

Graceful stop

close_agent

shutdown_request

TaskStop (lossy)

Redirect to different task

send_input

SendMessage

NO

Adjust scope mid-flight

send_input

SendMessage

NO

File conflict prevention

Manual git worktree routing

Native isolation: worktree + lead-only commits

None

Process isolation

YES (sub-process)

Shared worktree

Shared worktree

When to use each:

Scenario

Backend

Quick parallel tasks, coordination needed

Claude Native Teams

Codex-specific execution

Codex Sub-Agents

No team APIs available (last resort)

Background Tasks

Skill Invocation Across Runtimes

Skills that chain to other skills (e.g., /rpi calls /research, /vibe calls /council) MUST handle runtime differences:

Runtime

Tool

Behavior

Pattern

Claude Code

Skill(skill="X", args="...")

Executable — skill runs as a sub-invocation

Skill(skill="council", args="--quick validate recent")

Codex

N/A

Skills not available — inline the logic or skip

Check if Skill tool exists before calling

OpenCode

skill tool (read-only)

Load-only — returns <skill_content> blocks into context

Call skill(skill="council"), then follow the loaded instructions inline

OpenCode skill chaining rules:

• Call the skill tool to load the target skill's content into context

• Read and follow the loaded instructions directly — do NOT expect automatic execution

• NEVER use slashcommand syntax (e.g., /council) in OpenCode — it triggers a command lookup, not skill loading

• If the loaded skill references tools by Claude Code names, use OpenCode equivalents (see tool mapping below)

Cross-runtime tool mapping:

Claude Code

OpenCode

Notes

Task(subagent_type="...")

task(subagent_type="...")

Same semantics, different casing

Skill(skill="X")

skill tool (read-only)

Load content, then follow inline

AskUserQuestion

question

Same purpose, different name

TaskCreate, TaskUpdate, TaskList, TaskGet

todo

Task tracking (Claude uses 4 tools, OpenCode uses 1)

Read, Write, Edit, Bash, Glob, Grep

Same names

Identical across runtimes

Rules

• Never crash — missing CLI = skip or fallback, not error

• Always inform — tell the user what was skipped and how to enable it

• Preserve core function — the skill's primary purpose must still work without optional CLIs

• Progressive enhancement — CLIs add capabilities, their absence removes them cleanly

Reference Documents

• references/substring-rename-overreach.md — Pre-rename checklist for bulk sed across same-prefix concepts

• references/cross-harness-skill-parity.md — Knowledge parity beyond audit-codex-parity.sh; codex frontmatter strictness

• references/content-hash-cache.md

• references/compaction-signals.md

• references/backend-background-tasks.md

• references/backend-claude-teams.md

• references/backend-codex-subagents.md

• references/backend-inline.md

• references/claude-code-latest-features.md

• references/claude-cli-verified-commands.md

• references/codex-cli-verified-commands.md

• references/cli-command-failures-2026-02-26.md

• references/ralph-loop-contract.md

• references/orchestration-as-prompt.md

• references/stale-scope-validation.md — planning rule loaded by plan + pre-mortem: re-validate inherited scope against HEAD before acting on deferred beads or handoff docs.

• references/strict-delegation-contract.md — canonical contract loaded by /rpi, /discovery, /validation: strict sub-skill delegation is the default for top-level orchestrators.