/longrun

Run a long autonomous session that paces itself. Prepares scope. Checkpoints at phase boundaries. Closes with session log and retrospective. Three modes: prep, checkpoint, closeout.

Copy the command

/longrun

Defaults to the first mode below.

/longrun prep

Set up scope and gates before you start a long session.

/longrun checkpoint

Re-anchor mid-session — at a phase boundary, after compaction, or when scope drifts.

/longrun closeout

Wrap the session — write the session log, run the retro, bump whereami.

Type it bare, or paste it with loose context after — Claude figures out the rest.

Solves

Sessions over 50 turns drift. Context compacts and the goal disappears.

Payoff

A long autonomous session that paces itself. Prep and checkpoint and closeout — clean state every wake.

Takes in

prep or checkpoint or closeout mode plus scope hints

Returns

Goal doc with time budget
Checkpoint status at phase boundaries
Session log at closeout

Runs with

/value-prop /kiss /temperance /luminary /roadmap-reconcile /retro /promote /session-end /chronicle /pattern-review

Needs first

—

Called by

—

Deep dive — 15 sections · ~1,655 words · original SKILL.mdScroll for tutorial ↓

What it does

A long autonomous session that stays on track. You type /longrun prep once at session start. I set up scope, land the goal doc, touch gate markers, and start work. I checkpoint at every phase boundary, after each compaction, and every 25 steps. At close I write the session log, update whereami, run the retrospective, and open the PR. You get one clean landing when I finish — not a session that drifts mid-way.

Use /longrun when any of these hold:

Read the full skill

The complete instructions live at .claude/skills/longrun/SKILL.md in your repo after bassclef-sync. Open with your editor for phases, rules, and worked examples.

Where in the SDLC

Also relevant in: Concept, Design, Build, Test, Ship, Operate.

Why this phase: Multi-phase pacing composer — prepares scope, checkpoints at phase boundaries, closes with session log + retro. Home = Planning because that IS the composer's shape.

Skills catalog — browse all skills grouped by lifecycle phase
Tier system — what each tier gives you

FAQ

Terms from bassclef vocabulary that appear on this page.

luminary: A design or engineering thinker whose discipline bassclef borrows as a review lens — the practitioner in the cameo pattern above. Cooper for personas, Norman for signifiers, Feathers for characterization tests, Beck for TDD rhythm, plus about 90 others. Skills invoke a lens to check work against a specific discipline. `/luminary` browses the catalog. See: /docs/skills/luminary, doi:10.5281/zenodo.22145491
primitive: A building block of bassclef — a skill, a rule, a hook, a luminary, or an agent. Adopters compose primitives into workflows. The lite tier ships one primitive of each kind (except luminaries — lite ships 4, standard adds more, ultra loads the full 96). See: /docs/config/tier-system
composer: A bassclef skill that runs other skills as a chain. `/launch` composes `/interpret-input` + `/luminary` + `/stage` + `/spec` + `/decompose`. `/build` composes `/verify` + Builder dispatch. You invoke the composer; it invokes its parts. See: /docs/skills/launch, /docs/skills/build
temperance: The pause-and-think gate that fires before any non-trivial work. Two questions — am I building the right thing AND am I building it the right way? Bassclef enforces it with a pre-build hook that BLOCKs edits until a temperance marker exists on the branch. See: /docs/skills/temperance
characterization tests: Michael Feathers' term — tests that pin down what existing code ACTUALLY does (not what you wish it did) before you refactor. Bassclef adopts them at content-hash boundaries: the regenerator's SHA-256 check IS a characterization test — it pins the upstream body content so drift fails the build.
chronicle: A narrative log of what a session shipped — what worked, what did not, what got deferred. Written at session-end. Bassclef's chronicles live under `docs/chronicles/YYYY-MM-DD-slug.md`. Chronicles are how future sessions understand context that the code alone cannot show. See: /docs/skills/session-log
whereami: The authoritative project-state snapshot — current phase, active goals, subsystem status, gate progress, session recap history. Read at session-start; updated at session-end. The one file every new session should read first to know where the project stands. See: /docs/skills/whereami
hook: A shell script bassclef wires into Claude Code's tool-invocation lifecycle. Hooks fire at events like PreToolUse (before an Edit lands) or SessionStart (when a session opens). Hooks enforce rules mechanically — a rule says 'don't do X'; a hook fires and BLOCKs if X happens.
rule: A markdown file auto-loaded into every session's context. Rules pin discipline the agent applies to every response (plain-English writing, blocked-items protocol, temperance gates). Different from a hook — rules are steering; hooks are enforcement.
skill: A user-invocable command (like `/kiss` or `/temperance`) that runs a specific bassclef workflow. Every skill lives at `.claude/skills/<name>/SKILL.md` with frontmatter + procedure. Adopters type the slash command; Claude Code loads the skill and runs it.
adopter: A repo that has bassclef installed — either via `bassclef init` (lite tier from npm) or via the symlink pattern (standard tier with an Anthropic key). Bassclef's discipline: never break adopters. Every change to a downstream-observable surface ships with a compat shim.
bassclef-upstream: The private repo where bassclef substrate is authored. Ships to a public `bassclef` npm package + a public `bassclef-web` docs site. Adopters read the public tier; the operator maintains upstream.
session log: Same shape as a chronicle — narrative record of what a session shipped. `/session-log` skill drafts it from the session's commits + branch state. Feeds into `/journal` when the operator wants an audience-tuned draft. See: /docs/skills/session-log

Shipped by @kingofrock · Read the essays