**Executor resolution (host-aware):** - **Claude Code + `superpowers` installed** → prefer **`superpowers:brainstorming`** (its richer version, including an interactive visual companion). - **Otherwise** (Cursor / any host / no companion) → use this. Same discipline, portable (text-only — no v

The read side of the knowledge graph. Project memory grows past the context window; this pulls the *relevant slice* in before the work starts — retrieval by relevance, not recency — so the SDLC begins informed by history instead of re-deriving (or contradicting) what was already decided.

Setup, self-diagnosed. The powerful features each need a small step (a `gh` permission, a `pip install`, a filled north-star); this checks them all at once and hands you the exact fix, so nothing fails silently.

Run ONE goal through the full Goal-Based SDLC interactively, pausing for your approval at each gate, then record the outcome. Use when the user runs /sdlc-goal or wants to drive a single goal end-to-end with oversight.

**Optional. Only acts when `.sdlc/config.json` → `knowledge_graph.enabled` is `true`.** The graph accumulates two things, to keep enhancing the project's learnings:

Read side of the **local-only action log** (`skills/sdlc-loop/scripts/actionlog.py` writes it, opt-in via config `action_log.enabled`, default `false`) — a full-granularity trace of file touches, model/effort choices, subagent dispatch, and every mechanically-guaranteed loop action (claim/worktree

Run the autonomous park-and-continue SDLC loop over the .sdlc/goals backlog. Use when the user runs /sdlc-loop or asks to run goals autonomously / overnight / unattended.

Predict the model tier a goal deserves (haiku/sonnet/opus/fable) so a goal's phases run at a tier matched to the work. Use when the user runs /sdlc-model or asks which model to use for a goal.

**Executor resolution (host-aware):** - **Claude Code + `superpowers` installed** → prefer **`superpowers:writing-plans`** (its richer version). - **Otherwise** (Cursor / any host / no companion) → use this. Same discipline, portable.

The last gate before code. Review the active plan with two lenses; finish with one verdict.

A proactive research scout — the *supply* side that complements the gap log's *demand* side: it surfaces what you didn't know to look for. **Phase A is dry-run: it writes a digest under `.sdlc/` and records what it surfaced, but never files issues or touches GitHub.**

> The pre-flight checklist. Run before anything goes to prod.

Ground a sizing estimate in this repo's ACTUAL recent git throughput (commits/day, and merges-as-PRs/day) instead of "this feels like weeks". Converts a work estimate (N PR-sized units) into a calendar band at the measured rate. Use at the Plan phase, or when the user runs /sdlc-velocity or asks "how long will this take / what's our real pace".

**Executor resolution (host-aware):** - **Claude Code + `superpowers` installed** → prefer **`superpowers:verification-before-completion`** (its richer version). - **Otherwise** (Cursor / any host / no companion) → use this. Same discipline, portable.