agentleFS
mhlagharimhlaghari/stuntman/skills/scaffold/SKILL.md

scaffold

Skill16 starsChanged 2 months agoMITView on GitHub

Set up Stuntman project memory in AGENTS.md for Codex, CLAUDE.md for Claude Code, or both. Create and populate HANDOFF, STATUS, SPEC, and STRATEGY without clobbering existing content. Use when asked to scaffold project memory or set up handoff/status docs.

49 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use ascii-diagram, docs-ci and 10 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • SethGammonSkill

    ascii-diagram

    Generate perfectly aligned ASCII diagrams — architecture, flow, sequence, box-and-arrow. Uses a programmatic character-grid approach so alignment is guaranteed by math, not token prediction. Includes post-render verification.

    Also in 4 projects that use this

  • jh941213Skill

    docs-ci

    문서 CI(링크·OpenAPI lint·breaking 게이트·mermaid 검증)나 문서 드리프트 감지(docs.yaml 신선도 검사)를 설치하거나 돌릴 때 사용. 문서 내용 작성은 docs-architecture·docs-interfaces·docs-manuals.

    Also in 3 projects that use this

  • jh941213Skill

    docs-interfaces

    API·이벤트 인터페이스 문서(OpenAPI·swagger·AsyncAPI 명세, 엔드포인트 레퍼런스, API 구성도·시퀀스, API CHANGELOG)를 만들거나 코드와 맞출 때 사용. 시스템 구성도는 docs-architecture, API 설계 원칙은 api-design-principles.

    Also in 3 projects that use this

  • SethGammonSkill

    research-fleet

    Parallel research using Fleet wave mechanics. Spawns multiple scout agents, each investigating a different angle of the same question. Findings are compressed between waves. Produces a unified research brief from multiple independent perspectives.

    Also in 3 projects that use this

  • RachaelQuiselSkill

    client-docs

    Client documentation. Use when a build needs documentation updates or missing project documents need to be identified.

    Also in 3 projects that use this

  • vicoa-aiSkill

    live-preview

    Start a local app in the current user project, expose it through a tunnel, and return preview details in structured Markdown. Use when the user asks for a preview URL, public URL, shareable localhost, mobile/web preview, or tunnel setup.

    Also in 4 projects that use this

  • +4 more from jh941213

The file

We split this skill into the rules it gives an agent

3 rules. 0 are common, 1 is distinctive, and 2 are very specific to this project.

  • 0

    Common

    Most of the 42 edited versions we read keep these.

  • 1

    Distinctive

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Read host and tool setup. Set STUNTMAN_HOST to codex in Codex or claude in Claude Code.

    line 14
  2. Distinctive:

    Write all of them for a reader with zero memory of this session, and set each doc's date / changelog.

    line 69
  3. Very specific:

    Self-contained: no dependency on an external /checkpoint skill or memory system — just the host's instruction file + the living docs.

    line 89

“Common” is measured: kept by at least 6 in 10 of the 42 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

stuntman: scaffold — stand up the project-memory system

/delegate saves cost, /relay survives the rate limit; scaffold + handoff make a project document and resume itself across sessions. This command is the one-time setup; /handoff is the per-session resume.

The scaffolder

Read host and tool setup. Set STUNTMAN_HOST to codex in Codex or claude in Claude Code; use both when the user requests both hosts.

SCAFFOLD="$STUNTMAN_ROOT/bin/scaffold"
"$SCAFFOLD" --host "$STUNTMAN_HOST"

Idempotent and non-destructive. It:

  • writes a contract block (<!-- stuntman:scaffold:start … end -->) into the host's instruction file (AGENTS.md, CLAUDE.md, or both) — created if absent, appended if present, skipped if already there (recognizes the pre-0.5 stuntman:handoff marker too). Never rewrites existing memory or user-owned content.
  • appends a reply-style block (<!-- stuntman:style:start … end -->) to the same instruction file(s) — plain, short answers: answer first, five short sentences or fewer, one next step. Written once, so edits inside the block persist. To opt out, empty the block but keep its markers.
  • creates the living docs it references — HANDOFF.md (the session baton), STATUS.md (the board), SPEC.md (the contract), STRATEGY.md (the honest why) — each only if missing. Each self-declares as a living doc. HANDOFF.md and STATUS.md are snapshots, not logs: each session rewrites them (under 80 and 120 lines) and moves anything older to docs/handoff-archive.md.
  • appends or refreshes a managed agent roster block (<!-- stuntman:agents:start … end -->) in the same instruction file(s) — one full inventory snapshot shared by every selected host. Rerunning refreshes only that block; memory content is untouched, and user overrides persist outside the block. The roster states role preferences — Fable in Claude Code and Astra/GPT in Codex for spec writing and review, everything else executing through stunt — which are preferences, not availability or login proof. Installed means the CLI is on PATH, not that it is authenticated. Discovery runs opencode models / agy models with a short timeout and reads the local Codex model cache; it makes no model inference calls. Refreshing the snapshot never installs or configures models.

What to do

  1. Confirm the project root — run from the repo root (git root: git rev-parse --show-toplevel).
  2. Run the scaffolder and show the user its output.
  3. Populate the docs from reality. The stubs are living-doc skeletons — fill them in now so the system is useful immediately, by examining the project (its README.md, structure, recent git log):
    • SPEC.md — the contract: a one-paragraph Vision, the load-bearing Principles, Scope (in / out), Commands (exact build / test / run commands — run them or read them from the project's scripts, never guess), Boundaries (always / ask first / never), Success criteria (conditions a command or a person can check), Open decisions. Draft from what the project clearly is; flag guesses for the user.
    • STRATEGY.md — the honest Bottom line, Assessment, Direction. The critical version, not a pitch. If you can't assess honestly yet, say what you'd need to.
    • STATUS.md — the board: Built / In progress / Planned / Blockers.
    • HANDOFF.md — the baton: What changed this session, Verified (commands run and their results, what is unverified, committed or not), the Next step, Gotchas, dated. Write all of them for a reader with zero memory of this session, and set each doc's date / changelog. README.md is maintained too (refresh when the surface changes). Scaffold doesn't create one, so if it's missing — a brand-new project — infer what's being built from the code, structure, and any notes, then draft a README that states the purpose, how to run it, and the conventions the project should follow (best practices for its stack). Flag guesses for the user.
  4. Report: the system is active. Every future session now reads HANDOFF.md / STATUS.md first (the instruction lives in the host's instruction file, which auto-loads) and updates them before stopping. To resume any time, the user runs /handoff or says "execute handoff".

Notes

  • The contract is instruction-driven (the host's instruction file auto-loads, so read-first / update-before-stop is always in context), and a Stop hook (ships with the plugin) backs it up: in scaffolded projects only, if the session changed code but didn't update HANDOFF.md / STATUS.md, it nudges once before the turn ends. It fails open and never touches non-scaffolded projects. Codex only runs the hook after the user trusts it; the memory workflow does not require it.
  • Self-contained: no dependency on an external /checkpoint skill or memory system — just the host's instruction file + the living docs. (If the user already runs a /checkpoint that maintains STATUS.md, it stays compatible — same file.)
  • Pairs with /relay: when a long run pauses at the 5-hour cap, HANDOFF.md is the human-readable state the next session resumes from.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.