/resume

command: resume
disable-model-invocation: true
description: "[advanced] Resume a previous agent by ID — continue an interrupted task where it left off"

/octo:resume — Agent Resume

**Your first output line MUST be:** `🐙 Octopus Agent Resume`

Resume a previously-running Claude agent by ID. Picks up the agent's transcript and continues where it left off.

MANDATORY COMPLIANCE — DO NOT SKIP

**When the user explicitly invokes `/octo:resume`, you MUST call the `agent-resume` orchestrator path below.** You are PROHIBITED from pretending to resume an agent from memory or starting unrelated fresh work without telling the user.

Repository feature recovery

Before looking for a transcript or machine-local `.octo` state, inspect portable feature artifacts:

OCTO_ROOT="${CLAUDE_PLUGIN_ROOT:-${CODEX_PLUGIN_ROOT:-${HOME}/.claude-octopus/plugin}}"
python3 "$OCTO_ROOT/scripts/helpers/feature-contract.py" resume --root "$(pwd -P)" --explicit "<feature directory or spec path, empty when omitted>"

An unambiguous feature recovers intent, decisions, open markers and stable completed/pending task IDs from the repository. Historical completion is context and needs fresh verification. Missing raw runtime files are expected. If several unrelated features exist, ask the user to select one; never choose by recency. Pass the selected feature to the existing `agent-resume` orchestrator path. Explicit Claude agent IDs retain transcript continuation.

A feature with a spec only resumes planning. A feature with plan/tasks resumes pending implementation through the same policy, clarification, analysis and wave-validation gates. Run the returned decision batch once through the host's native question tool before the next plan/develop boundary. Keep skipped decisions open.

Step 1: Get the Agent ID

If you don't have the agent ID:

• Check `/octo:sentinel` output for running agent IDs
• Look in `~/.claude-octopus/results/` for recent result files (filename prefix contains agent type + task ID)
• The agent ID was shown when the agent was originally spawned

Step 2: Resume

Use the Bash tool to execute:

**Preflight check — Ensure plugin root is resolvable (run via Bash tool FIRST):**

set -euo pipefail

OCTO_ROOT="${CLAUDE_PLUGIN_ROOT:-${CODEX_PLUGIN_ROOT:-${HOME}/.claude-octopus/plugin}}"
if [[ ! -x "$OCTO_ROOT/scripts/orchestrate.sh" ]]; then
  helper="$OCTO_ROOT/scripts/helpers/ensure-plugin-root.sh"
  if [[ ! -x "$helper" ]]; then
    helper="$(find "${HOME}/.claude/plugins/cache" "${HOME}/Library/Application Support/Claude" "${LOCALAPPDATA:-/dev/null}/Claude" "${XDG_DATA_HOME:-${HOME}/.local/share}/Claude" -maxdepth 8 -path "*/nyldn-plugins/octo/*/scripts/helpers/ensure-plugin-root.sh" -print -quit 2>/dev/null)"
  fi
  [[ -x "$helper" ]] && bash "$helper" >/dev/null 2>&1 || true
fi
test -x "$OCTO_ROOT/scripts/orchestrate.sh" && echo "plugin-root:ok" || echo "plugin-root:missing"

If the output is `plugin-root:missing`, stop and ask the user to run `/octo:setup`.

"${CLAUDE_PLUGIN_ROOT:-${CODEX_PLUGIN_ROOT:-${HOME}/.claude-octopus/plugin}}/scripts/orchestrate.sh" agent-resume "$ARGUMENTS"

Pass the agent ID as `$ARGUMENTS`. Optionally append a follow-up prompt:

# Just agent ID (resumes with "Continue where you left off.")
orchestrate.sh agent-resume abc123

# Agent ID + custom prompt
orchestrate.sh agent-resume abc123 "fix the failing test in auth.ts"

Requirements

• Claude Code v2.1.34+ (`SUPPORTS_CONTINUATION=true`, `SUPPORTS_STABLE_AGENT_TEAMS=true`)
• Agent Teams enabled (required for agent transcript access)
• Agent must have been a Claude agent (not an external CLI agent — those don't support transcripts)
• CC v2.1.77+: Resume uses `SendMessage` (auto-resumes stopped agents). The `Agent(resume:)` parameter was removed in v2.1.77.

Run `/octo:skill-doctor` to verify that the required flags are active.

Fallback

If continuation is not supported, spawn a fresh agent with `/octo:develop` and describe what was in progress.