원클릭으로 Manus에서 모든 스킬 실행

eval-spec-vs-build

스타2

포크2

업데이트2026년 6월 18일 10:41

Scores how faithfully a recorded build run followed its spec.md, in the spec-build-lab. Use this skill ANY time the user asks to evaluate, score, check, judge, assess, or compare a specific build run AGAINST THE SPEC — including phrasings like "evaluate run-02 against the spec", "score run-02 against the spec", "did the build follow the spec", "did run-02 implement the spec", "spec fidelity check on run-NN", "how closely did run-02 follow spec.md", "run a spec-vs-build eval", "evaluate the spec adherence of the latest run", or "/eval-spec-vs-build run-NN". Reads the per-run snapshot evals/<run>/spec.md (NOT the live spec.md) and the contents of builds/<run>/, then writes evals/<run>/spec-vs-build.md. Requires an explicit run-NN; if none is given, list the runs in builds/ and ASK which one — never default. Do NOT use this skill to evaluate against course materials — that's /eval-materials-vs-build, a different reference and different question.

설치

Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.

Manus에서 실행

출처

https-deeplearning-ai

https-deeplearning-ai/spec-build-lab

GitHub 저장소 열기 Creator 저장소 보기

다운로드

Manus에서 실행

eval-spec-vs-build

Compare a recorded run's app against the spec snapshot it was built from. Spec fidelity only. Course alignment is /eval-materials-vs-build.

Steps

Verify cwd is a course folder. Confirm spec.md, materials/, builds/, evals/ exist. Refuse otherwise: "Run /eval-spec-vs-build from inside a course folder."
Require a run. If $ARGUMENTS is empty:
```
ls -1 builds/ 2>/dev/null | grep -E '^run-[0-9]+$' | sort
```
List those runs and ask the user: "Which run should I evaluate? Pick one — I won't default to the latest." Stop until they answer.
Validate the run. Confirm evals/$ARGUMENTS/spec.md (the snapshot) exists and builds/$ARGUMENTS/ is a non-empty directory. Refuse otherwise. The report reflects whatever the build folder contains — evaluating an in-progress run is fine; just re-run after the build finishes for a clean score.
Read the inputs.
- Snapshot spec: evals/$ARGUMENTS/spec.md (NOT the top-level spec.md — that may have been regenerated since this run).
- Build: walk builds/$ARGUMENTS/ (skip node_modules, .venv, dist, build, .git, and anything matching its own .gitignore).
Produce the report at evals/$ARGUMENTS/spec-vs-build.md with sections:
- Summary — one-paragraph verdict.
- Coverage matrix — each spec section → present / partial / missing, with file references into builds/$ARGUMENTS/.
- Divergences — things the build did that the spec did not specify.
- Missing — things the spec specified that the build did not do.
- Score — your judgment on a 0–10 scale, with reasoning.
Report the path written, then hand off. After echoing the report path, tell the user: "Spec-fidelity report written. The companion eval is /eval-materials-vs-build $ARGUMENTS — different reference, different question (course alignment, not spec fidelity). Say 'go ahead' to run it next, or skip if you've already done it."

Don't

Don't read materials/ — that's the other eval. Mixing them defeats the point of having two.
Don't read the top-level spec.md — always the per-run snapshot.
Don't pick a default run. List the runs in builds/ and ask the user which one.

이 저장소의 다른 Skills

같은 저장소

generate-spec

https-deeplearning-ai/spec-build-lab

Produces spec.md for a course in the spec-build-lab by reading the course's materials and writing a build-ready specification. Use this skill ANY time the user asks to generate, write, produce, build, regenerate, create, or run the spec for a course — including phrasings like "generate the spec", "write spec.md", "produce spec.md", "make a spec from this course", "turn the materials into a spec", "build the spec out of materials/", "regenerate spec.md", "run the spec generation", "I've added the notebooks, please create the spec", or "/generate-spec". Optionally accepts a git repo (SSH URL or local path) as an argument to auto-ingest the notebooks + helper.py into materials/notebooks/ before generating; without an argument it uses whatever is already in materials/. Reads every file under materials/notebooks/ and materials/transcripts/ and follows references/spec-generation-guide.md. Writes spec.md in the current working directory. Must run from inside a course folder (courses/<name>/). Do NOT use this skill f

2026-07-202

new-course

https-deeplearning-ai/spec-build-lab

Adds a new course to the spec-build-lab. Use this skill ANY time the user wants to register, scaffold, set up, add, create, init, or start tracking a NAMED COURSE in this repo — including phrasings like "scaffold a new course called X", "set up a course folder for X", "make a new course called X", "add the X course", "init the X course", "let's start tracking the X course", "I want to add a new course", "/new-course X", or any user message naming a course they want to begin. This skill copies templates/course/ to courses/<name>/ and nothing else; it does NOT read materials, write spec.md, or run a build. If the user asks anything about an EXISTING course, do NOT use this skill. Always run from the repo root.

2026-07-012

eval-materials-vs-build

https-deeplearning-ai/spec-build-lab

Scores how faithfully a recorded build run aligned with the course materials, in the spec-build-lab. Use this skill ANY time the user asks to evaluate, score, check, judge, or compare a specific build run AGAINST THE COURSE — including phrasings like "evaluate run-02 against the materials", "score run-02 against the course", "is run-02 faithful to the course", "did the build match the course", "course alignment check on run-NN", "how does run-02 compare to the notebooks and transcripts", "is the build aligned with what the course taught", "run a materials-vs-build eval", or "/eval-materials-vs-build run-NN". Reads materials/notebooks/, materials/transcripts/, and builds/<run>/, then writes evals/<run>/materials-vs-build.md. Requires an explicit run-NN; if none is given, list the runs in builds/ and ASK which one — never default. Do NOT use this skill to evaluate against spec.md — that's /eval-spec-vs-build, a different reference and different question.

2026-06-182

extract-build-log

https-deeplearning-ai/spec-build-lab

Extracts the build conversation for a recorded run in the spec-build-lab and writes it to evals/<run>/session-log.md. Use this skill ANY time the user wants to capture, save, snapshot, extract, export, or render the conversation for a specific build run — including phrasings like "extract the build log", "save the conversation for run-02", "snapshot run-02", "export the build log", "render the session log", "log this build", "produce the session log for run-NN", or "/extract-build-log run-NN". Reads the transcript recorded by Claude Code (located via the .session breadcrumb at evals/<run>/.session) and slices it between an opening bookend (the /prepare-build announcement) and a closing bookend (a fuzzy diagram/structure phrase by default, or a user-supplied --until pattern). Idempotent — re-runnable with different bookends. Requires an explicit run-NN; if none is given, list the runs in builds/ and ASK which one — never default. Do NOT use this skill to evaluate the run — that's /eval-spec-vs-build or /eval-m

2026-06-182

prepare-build

https-deeplearning-ai/spec-build-lab

Prepares a new build run for a course in the spec-build-lab — allocates the run folders, snapshots the spec into a frozen eval copy and a working copy, drops a session breadcrumb, and moves into the build folder. Does NOT do the building itself; that's the user + agent's subsequent work in builds/run-NN/. Use this skill ANY time the user signals intent to begin building an app from spec.md — including phrasings like "prepare a build", "set up a build", "start a new build", "begin the build", "kick off the build", "kick off run-02", "let's build the app", "let's build it", "build the X app from the spec", "implement the spec", "I'm ready, let's build", "let's do another build", "/prepare-build", or simply launching into coding work inside a course folder that has a spec.md. INVOKE PROACTIVELY — treat any clear build intent as the trigger. Do NOT use this skill to run an existing build, build a docker image, or build features in unrelated projects.

2026-06-182

name	eval-spec-vs-build
description	Scores how faithfully a recorded build run followed its spec.md, in the spec-build-lab. Use this skill ANY time the user asks to evaluate, score, check, judge, assess, or compare a specific build run AGAINST THE SPEC — including phrasings like "evaluate run-02 against the spec", "score run-02 against the spec", "did the build follow the spec", "did run-02 implement the spec", "spec fidelity check on run-NN", "how closely did run-02 follow spec.md", "run a spec-vs-build eval", "evaluate the spec adherence of the latest run", or "/eval-spec-vs-build run-NN". Reads the per-run snapshot evals/<run>/spec.md (NOT the live spec.md) and the contents of builds/<run>/, then writes evals/<run>/spec-vs-build.md. Requires an explicit run-NN; if none is given, list the runs in builds/ and ASK which one — never default. Do NOT use this skill to evaluate against course materials — that's /eval-materials-vs-build, a different reference and different question.
argument-hint	<run-NN>
allowed-tools	["Read","Write","Glob","Grep","Bash"]

eval-spec-vs-build

Compare a recorded run's app against the spec snapshot it was built from. Spec fidelity only. Course alignment is /eval-materials-vs-build.

Steps

Verify cwd is a course folder. Confirm spec.md, materials/, builds/, evals/ exist. Refuse otherwise: "Run /eval-spec-vs-build from inside a course folder."
Require a run. If $ARGUMENTS is empty:
```
ls -1 builds/ 2>/dev/null | grep -E '^run-[0-9]+$' | sort
```
List those runs and ask the user: "Which run should I evaluate? Pick one — I won't default to the latest." Stop until they answer.
Validate the run. Confirm evals/$ARGUMENTS/spec.md (the snapshot) exists and builds/$ARGUMENTS/ is a non-empty directory. Refuse otherwise. The report reflects whatever the build folder contains — evaluating an in-progress run is fine; just re-run after the build finishes for a clean score.
Read the inputs.
- Snapshot spec: evals/$ARGUMENTS/spec.md (NOT the top-level spec.md — that may have been regenerated since this run).
- Build: walk builds/$ARGUMENTS/ (skip node_modules, .venv, dist, build, .git, and anything matching its own .gitignore).
Produce the report at evals/$ARGUMENTS/spec-vs-build.md with sections:
- Summary — one-paragraph verdict.
- Coverage matrix — each spec section → present / partial / missing, with file references into builds/$ARGUMENTS/.
- Divergences — things the build did that the spec did not specify.
- Missing — things the spec specified that the build did not do.
- Score — your judgment on a 0–10 scale, with reasoning.
Report the path written, then hand off. After echoing the report path, tell the user: "Spec-fidelity report written. The companion eval is /eval-materials-vs-build $ARGUMENTS — different reference, different question (course alignment, not spec fidelity). Say 'go ahead' to run it next, or skip if you've already done it."

Don't

Don't read materials/ — that's the other eval. Mixing them defeats the point of having two.
Don't read the top-level spec.md — always the per-run snapshot.
Don't pick a default run. List the runs in builds/ and ask the user which one.