Plan orchestration for Claude Code: subagents scout, implement, and verify each part of a markdown plan, with one commit per part and state that survives sessions.
Agent
Write-only, headless agent that reads a captured-answers scratch file produced by /su:discover and writes a single Suhail-format plan file. Sole writer of the plan — it does not interview the user, run commands, or modify anything other than the output plan file. Invoked only by /su:discover (Phase 5) after the…
Agent
One-shot read-only scanner that grounds /su:discover Phase 0. Dispatched once before the interview begins. Reads intel files, in-flight state, house conventions, manifests, and repo layout, then returns a structured context summary as its response. Writes nothing.
Agent
Implements one Part by writing code per the scout's step list. Generic — runs whatever commands the brief and steps specify. Writes a summary to .suhail/parts/ /execution.md. Never commits. Never deploys. Both are flagged as Manual follow-ups. Invoked only by the suhail orchestrator.
Agent
One-shot project scanner. Reads project manifests, conventions docs, and directory layout, then writes a structured intel cache under .suhail/intel/. Stack-agnostic — discovers conventions on its own. Invoked only by /su:init.
Agent
Merged researcher+planner role in the Suhail pipeline. Given a Part description, reads the intel directory, explores the codebase for files-to-touch, gotchas, and reusable helpers, then drafts an ordered step list — all in one pass. Writes exactly one file: brief.md at the path the orchestrator provides. Invoked only…
Agent
Merged reviewer + security-auditor role. Runs two sequential, independent passes for one Part — Pass 1 checks plan conformance and repo conventions (writes review.md), Pass 2 audits for security risk (writes audit.md). Invoked only by the suhail orchestrator.