# Engineering

Skills I use daily for code work.

- **[afk](./afk/SKILL.md)** — Autonomous loop that drains the `ready-for-agent` queue. Claims an issue, runs it in an isolated worktree via claude/codex, merges back to main, closes. Filters by Spec or explicit issue list; pins caller identity with `RED_AFK_RUNNER`; swaps runners only when explicitly requested; heartbeat + monitor + live progress.
- **[ask-red](./ask-red/SKILL.md)** — Router over RedSkills flows. Answers "what do I do now?" by mapping situations to `/afk`, `/go`, the idea path, HITL/retake, diagnosis, reporting, and maintenance routes.
- **[guard-process-birth](./guard-process-birth/SKILL.md)** — Path-declared guard brief that keeps Worker process birth behind the host-scoped `redskilled` daemon when per-project runtime source is touched.
- **[guard-serialization](./guard-serialization/SKILL.md)** — Path-declared guard brief that keeps structured files, append streams, and owned wires on TOON/TOONL when app or package source is touched.
- **[go](./go/SKILL.md)** — Semi-structured front door between `/goal` and `/afk`. Mints a disposable tracking issue in the isolated `lane:go` lane (out of `ready-for-agent`, so a running drain cannot claim it), spins a Worker under the shared `.red/tmp/workers/` root with `origin=go` and `current.kind=go`, reuses the whole AFK engine with the interactive gate, and brings back a PR that auto-closes the issue on merge. Works with or without a drain running.
- **[manager](./manager/SKILL.md)** — Operator liaison for one effort. `$dev:manager <intent>` starts an effort and persists it in the operator-and-host-scoped portfolio (`~/.red/manager/efforts/<id>.toonl`, versioned schema, generation counter, secret-free); `manager status [id]` renders its brief from owned lifecycle state.
- **[wayfinder](./wayfinder/SKILL.md)** — Planning on-ramp for work too large for one agent session. Creates one tracker map (`wayfinder:map`) with destination, in-scope fog, and out-of-scope boundaries, then publishes one-session research/grilling/prototype/task children into the standard AFK/HITL queues.
- **[dashboard](./dashboard/SKILL.md)** — RedSkills process dashboard: open Specs/issues, global and local AFK workers, flow metrics, and DORA proxies.
- **[audit-skills](./audit-skills/SKILL.md)** — Read-only skill-quality auditor. Scores every shipped SKILL.md against the house style (mechanical lint checks + an LLM judge on the dev review engine), overlays best-effort telemetry, and prints a worst-first scorecard. Zero side effects — no `--fix`, no git, no gh.
- **[daily-review](./daily-review/SKILL.md)** — Period-parameterized operational review (`--period day` default; `--period week` for six days): delivery big numbers, local worker attempts/time, tokens when available, HITL/blocker challenges, and issue/PR cycle times.
- **[model-tier-policy](./model-tier-policy/SKILL.md)** — Cross-host model tier policy for Claude Code, Codex, and AFK: tier table, deterministic-first validation, simple-vs-complex classification, escalation, and executor/config pointers.
- **[retake](./retake/SKILL.md)** — Diagnose-then-act issue recovery. Reconstructs the issue's full state (linked PRs, matching branches, local worktrees, uncommitted work, HITL state, current blocker), reports where the work stands, then executes the right action: a plain requeue, an `--adopt-branch` landing through the no-agent gate, or a handoff to `/hitl`.
- **[curate](./curate/SKILL.md)** — Interactive, archive-only Skill curator. Lists `archive` candidates from `memory curate skills --json`, requires explicit approval, archives approved Curatable skills (recoverable atomic move + SHA-256 manifest), and reverses with `/curate --restore <name>`. Tracer slice for the mutating curator.
- **[context](./context/SKILL.md)** — Compose the RedSkills context stack before non-trivial work: domain docs, ADRs, LLM Wiki, Memory graph/recall, graph-aware zoom-out, and self-improvement telemetry.
- **[how](./how/SKILL.md)** — Explain how a subsystem works with parallel explorers and a synthesizing explainer; on request, critique its architecture with an independent panel.
- **[diagnose](./diagnose/SKILL.md)** — Disciplined diagnosis loop for hard bugs and performance regressions: reproduce → minimise → hypothesise → instrument → fix → regression-test.
- **[ground-truth](./ground-truth/SKILL.md)** — Adversarial self-verification: claim → fresh ground-truth → confirm. Re-derive a state change from a fresh a11y-tree/DB/file/test read (browser vehicle: chrome-devtools-axi, numbered refs + stale-ref validation + TOON) before reporting success. Evidence standard for `/verify`, `/code-review`, and the frontend skills.
- **[start](./start/SKILL.md)** — Grilling session that challenges your plan against the existing domain model, sharpens terminology, and updates `.red/CONTEXT.md` and ADRs inline.
- **[hitl](./hitl/SKILL.md)** — Resolve one `ready-for-human` issue by extracting the pending human decision, recording the answer as Human guidance, and promoting back to `ready-for-agent` when delegable.
- **[triage](./triage/SKILL.md)** — Triage issues through a state machine of triage roles.
- **[report-bug](./report-bug/SKILL.md)** — Interview the user about a bug they hit, then file a `type:bug needs-triage` issue on the project tracker. Seeds from conversation context when invoked with no argument.
- **[improve-codebase-architecture](./improve-codebase-architecture/SKILL.md)** — Find deepening opportunities in a codebase, informed by the domain language in `.red/CONTEXT.md` and the decisions in `.red/adr/`.
- **[red-setup](./red-setup/SKILL.md)** — Scaffold the per-repo config (issue tracker, triage label vocabulary, domain doc layout) that the other engineering skills consume.
- **[redskilled](./redskilled/SKILL.md)** — Operate the per-machine execution daemon: inspect host state, provision its home, configure host ceilings, restart it without evacuating Workers, and confirm setting origins. Given an issue number, it instead writes a self-serve debug dossier for that issue's Worker(s).
- **[red-doctor](./red-doctor/SKILL.md)** — Adoption/process doctor. Reports stack drift plus the shared operational probes that fleet boot also consumes (label vocabulary, AGENTS≡CLAUDE parity, statusline form, MCP wiring, `blocked:*` hygiene, boot-refusal probes) and names each fix's home. Read-only by default; `--fix` applies only the approved, per-finding gated repair path. The recurring counterpart to `/red-setup`.
- **[red-gains](./red-gains/SKILL.md)** — Detailed rsp usage-gains report: latency percentiles, throughput activity, token-savings trends, command-family winners, and degradation health from the shared RedDB telemetry store.
- **[adr-editor](./adr-editor/SKILL.md)** — Totipotent editor over the `.red/adr/` decision collection. Opens by asking which subject to work on, offering the collection's real INDEX themes and title-term clusters with counts, then answers about that slice: lists and groups the records by subject, surfaces inconsistencies (contradictions, overlaps, stale paths, dangling supersede pointers, INDEX drift), and applies **every** mutation in-session: add, remove, rewrite, merge, split, archive, renumber, re-index. The maintainer decides, the skill executes; one confirmation gates a destructive or wide batch.
- **[red-statusline](./red-statusline/SKILL.md)** — Install or inspect the RedSkills statusline for the active host: Claude Code command-backed AFK block, Codex native `tui.status_line` footer plus self-heal hook.
- **[implement](./implement/SKILL.md)** — Interactive, human-driven Spec execution: guided `/tdd` at pre-agreed seams, typechecking, full-suite gate, `/code-review`, and `/retake` to land. The interactive counterpart to the autonomous `/afk` fleet.
- **[tdd](./tdd/SKILL.md)** — Test-driven development with a red → green loop. Builds features or fixes bugs one vertical slice at a time, at pre-agreed seams.
- **[to-tickets](./to-tickets/SKILL.md)** — Break any plan or Spec into independently-grabbable Tickets (GitHub issues) using vertical slices.
- **[to-spec](./to-spec/SKILL.md)** — Turn the current conversation context into a Spec and submit it as a GitHub issue.
- **[zoom-out](./zoom-out/SKILL.md)** — Map-first Codebase understanding for unfamiliar code; graph-aware when Memory Graph mode is ready and read-only when it is not.
- **[prototype](./prototype/SKILL.md)** — Build a throwaway prototype to flesh out a design — either a runnable terminal app for state/business-logic questions, or several radically different UI variations toggleable from one route.
- **[wizard](./wizard/SKILL.md)** — Generate an interactive Bash wizard for credentials, CI secrets, third-party dashboards, provisioning, migrations, or cutovers that need human action.
- **[verify](./verify/SKILL.md)** — Verify that a code change does what it's supposed to by driving a live app with the CDP driver and checking ground-truth snapshots. Every visible-state claim must be backed by a `red-browser snapshot` result; no snapshot, no success claim.
- **[code-review](./code-review/SKILL.md)** — Two-axis diff review (Standards + Spec) via parallel sub-agents. Standards axis includes the always-on Fowler smell baseline; Spec axis verifies the diff against the originating issue/Spec.
- **[resolving-merge-conflicts](./resolving-merge-conflicts/SKILL.md)** — Guided merge-conflict resolution: inspect state, find each side's original intent, resolve every hunk preserving both intents, then run the project's automated checks. Never abort.