/debug

command: debug
disable-model-invocation: true
description: Debug a symptom through reproducible, bounded evidence

Octopus debug

For long Claude investigations, follow the subagent context and effort guidance in `docs/CLAUDE-NATIVE-INTEGRATION.md`. An advisor requires explicit activation and adds usage beyond the current worker. Preserve host-only and budget limits.

Load `skills/blocks/engineering-method-selection.md` from the installed plugin and apply only the methods relevant to this task. Preserve this entry point's execution contract and output format. Read referenced skills as instructions; do not invoke the current command recursively or add provider calls from a seat.

Load and follow `${CLAUDE_PLUGIN_ROOT:-${CODEX_PLUGIN_ROOT:-${HOME}/.claude-octopus/plugin}}/.claude/skills/skill-debug/SKILL.md` and its literal `skills/blocks/debug-feedback-loop.md` reference.

Treat `--peer-review` as an instruction to request one bounded independent review through existing Octopus routing. Never interpolate the flag or user text into a shell command. Natural-language independent-review requests have the same meaning. Otherwise run on the current host with zero additional provider dispatches unless an existing escalation policy both requires and permits review under the effective preferences and billing limits. Risk alone does not authorize a paid call; honor explicit host-only requests.

Finish with the structured reproduction record. `complete` requires the original scenario to pass and temporary instrumentation to be removed. Use `inconclusive` when the required environment was unavailable.