From idc
Use when an IDC role running under Claude Code must turn an abstract runtime primitive (durable worker / bounded fan-out / goal loop) into concrete Claude mechanics, or resolve a model tier to a concrete model.
How this skill is triggered — by the user, by Claude, or both
Slash command
/idc:idc-adapter-claudeThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
The Claude Code runtime adapter. IDC process docs are written against three abstract
The Claude Code runtime adapter. IDC process docs are written against three abstract
primitives (WORKFLOW.md §5); this skill maps each to concrete Claude Code mechanics and
resolves the tier-symbolic model table (WORKFLOW.md §6) to concrete models at spawn time.
There is exactly one adapter per runtime — this is the only place Claude mechanics live, so
the process docs stay runtime-neutral and cannot drift per-runtime.
| Primitive | Claude Code mechanic | Fallback |
|---|---|---|
| Durable worker (Build implementer, autorun lane) | A Claude Teams teammate in its own cmux pane: the lead pre-creates a worktree, TeamCreates, then spawns the teammate with Agent({team_name, …}) pointed at that worktree (NOT the isolation:"worktree" param, which silently runs on main); coordinate via SendMessage; tear down with TeamDelete. | No teams environment → run the work serially in the main session, one unit at a time. |
| Bounded fan-out (domain experts, drafters, clash pairs, reviewers) | The Workflow tool for deterministic fan-out/pipeline, or parallel Agent (Task) subagents for independent reads. Review fan-out is always fresh-context subagents (cold read = adversarial independence, token-optimal). | Subagents are available in every Claude environment; no fallback needed. |
| Goal loop (issue execution) | The native /goal loop with auto-goal discipline: render-before-run, record-and-vary iteration, evidence-before-assertion, the attempt ceiling, and the no-punt rule. The issue body IS the contract. | — |
git worktree add .claude/worktrees/<name> -b worktree-<name>)
and verify with git worktree list before any worker writes; the teammate pins git to
the worktree (git -C <path> or cd first).Read WORKFLOW-config.yaml::model_routing. For a spawn at tier <tier>, resolve
model_routing.<tier>.claude.model and apply its thinking / effort hint. Process docs
name only the tier; this skill resolves the concrete model. The Recirculator maintains the table when
models change — never hardcode a model id in a command, agent, or non-adapter skill.
WORKFLOW-config.yaml also carries the gating: requirements-gate toggle (gating.prd /
gating.trd), but that is read by the gate predicate (scripts/idc_recirculator_layers.py) for
Plan and the Recirculator — not by tier resolution.
| Tier | Resolves to (per config) | Applied to |
|---|---|---|
reasoning | model_routing.reasoning.claude | planning cognition; review coordinator/verdict + judgment dimensions; recirculation analysis + PRD diffs; clash/matrix + sequencing; merge deconfliction |
standard | model_routing.standard.claude | think/interview; build implementers; finisher/orchestrator; autorun parent |
utility | model_routing.utility.claude | execute-never-decide: research digestion, recon, templated emission, board mechanics, the schema check, inventory review dimensions |
WORKFLOW-config.yaml (the Recirculator owns the
model table).npx claudepluginhub llamallamaredpajama/idc-workflow --plugin idcGuides creation, editing, and verification of skills for AI coding agents using test-driven development with subagent scenarios. Use when authoring or debugging skills.