handoff

Name: handoff
Author: arimet

27stars0forks

How to install handoff?

arimet/handoff/handoff

$npx -y skills add arimet/handoff --skill handoff

Installs into the current project.

Summary

Use when a feature or code change made by Claude is complete and the developer wants to confirm they actually understood it before shipping. Triggers on "handoff", "quiz me", "comprehension check", "did I understand this", "verify I read the code", "test my understanding", or aft

Files · 1

View on GitHub

SKILL.md

1	# Handoff — verify you understood Claude's code
2
3	When Claude builds a feature, it is easy to merge the result without truly
4	reading or understanding it. `handoff` turns passive review into an active
5	check: it interviews the developer to confirm they read the code (where the
6	logic lives) and understood it (why decisions were made, how it behaves),
7	then reports a score and the spots to revisit.
8
9	This skill conducts an interactive quiz. Ask questions one at a time and
10	wait for an answer before continuing. Conduct the entire quiz in the
11	developer's language (match the language they are writing in).
12
13	## Step 1 — Gather the material (hybrid source)
14
15	Build the quiz from two complementary sources:
16
17	1. Conversation context — what was built this session: the files touched,
18	the key decisions and trade-offs, what was deliberately NOT done, and what
19	could plausibly break. This is the source for the "why".
20	2. Git diff (when available) — to objectively anchor "where" and check that
21	nothing important is missed. Run:
22	- `git rev-parse --git-dir` to confirm a repo exists.
23	- `git diff --stat` then `git diff` (covers unstaged + staged) to enumerate
24	changed files and lines. Use real `file:line` locations from this diff.
25
26	If there is no git repo, degrade gracefully and rely on conversation context
27	alone — still produce a useful quiz.
28
29	## Step 2 — Calibrate the number of questions
30
31	Scale to the size of the change (use the diff stat, or your best estimate from
32	context if there is no git):
33
34	\| Change size \| Questions \|
35	\| ---------------------------------------- \| --------- \|
36	\| Small (< ~50 lines, or 1–2 files) \| 3 \|
37	\| Medium \| 5 \|
38	\| Large (> ~300 lines, or ≥ 5 files) \| 8–10 \|
39
40	## Step 3 — Generate a mixed question set
41
42	Cover both reading and understanding. Mix these types:
43
44	- Where (multiple choice). "Which file/function contains X?" Provide
45	plausible distractors drawn from other real files in the change.
46	- Why (open-ended). "Why was approach X chosen over Y?", "What trade-off
47	does this make?", "What would have broken if we hadn't done Z?"
48	- Behavior (multiple choice or open). "If input X arrives, what happens?",
49	"How would you debug it if X fails?"
50
51	Favor the parts of the change that matter most (core logic, tricky edge cases,
52	non-obvious decisions) over trivial boilerplate.
53
54	## Step 4 — Ask one question at a time
55
56	Present a single question, wait for the answer, then evaluate it before moving
57	on. Never dump the whole quiz at once.
58
59	## Step 5 — Evaluate each answer
60
61	- Correct → confirm briefly and move to the next question.
62	- Wrong or incomplete → point to the relevant `file:line` as a hint
63	without revealing the answer, and invite a second attempt.
64	- If the second attempt is still wrong → reveal the correct answer with a
65	short explanation, then continue.
66
67	Track the result of each question (correct / correct-on-retry / missed) to
68	build the final score.
69
70	## Step 6 — Final score and weak areas
71
72	End with:
73
74	- A score, e.g. `Score: 4/6`.
75	- The weak areas: the topics missed, each with a `file:line` pointer to
76	re-read. Keep the tone encouraging, not punitive — the goal is to know what
77	to revisit, not to grade harshly.
78
79	## Optional: automatic suggestion (Stop hook)
80
81	This skill ships an optional, non-blocking Stop hook that suggests running
82	`/handoff` when a session produced a large diff. It never forces the quiz. See
83	[references/install.md](references/install.md) for setup and thresholds.

People who liked this also viewed

6 picks

Type

handoff

27stars0forks

View on GitHub

How to install handoff?

arimet/handoff/handoff

$npx -y skills add arimet/handoff --skill handoff

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# Handoff — verify you understood Claude's code
2
3	When Claude builds a feature, it is easy to merge the result without truly
4	reading or understanding it. `handoff` turns passive review into an active
5	check: it interviews the developer to confirm they read the code (where the
6	logic lives) and understood it (why decisions were made, how it behaves),
7	then reports a score and the spots to revisit.
8
9	This skill conducts an interactive quiz. Ask questions one at a time and
10	wait for an answer before continuing. Conduct the entire quiz in the
11	developer's language (match the language they are writing in).
12
13	## Step 1 — Gather the material (hybrid source)
14
15	Build the quiz from two complementary sources:
16
17	1. Conversation context — what was built this session: the files touched,
18	the key decisions and trade-offs, what was deliberately NOT done, and what
19	could plausibly break. This is the source for the "why".
20	2. Git diff (when available) — to objectively anchor "where" and check that
21	nothing important is missed. Run:
22	- `git rev-parse --git-dir` to confirm a repo exists.
23	- `git diff --stat` then `git diff` (covers unstaged + staged) to enumerate
24	changed files and lines. Use real `file:line` locations from this diff.
25
26	If there is no git repo, degrade gracefully and rely on conversation context
27	alone — still produce a useful quiz.
28
29	## Step 2 — Calibrate the number of questions
30
31	Scale to the size of the change (use the diff stat, or your best estimate from
32	context if there is no git):
33
34	\| Change size \| Questions \|
35	\| ---------------------------------------- \| --------- \|
36	\| Small (< ~50 lines, or 1–2 files) \| 3 \|
37	\| Medium \| 5 \|
38	\| Large (> ~300 lines, or ≥ 5 files) \| 8–10 \|
39
40	## Step 3 — Generate a mixed question set
41
42	Cover both reading and understanding. Mix these types:
43
44	- Where (multiple choice). "Which file/function contains X?" Provide
45	plausible distractors drawn from other real files in the change.
46	- Why (open-ended). "Why was approach X chosen over Y?", "What trade-off
47	does this make?", "What would have broken if we hadn't done Z?"
48	- Behavior (multiple choice or open). "If input X arrives, what happens?",
49	"How would you debug it if X fails?"
50
51	Favor the parts of the change that matter most (core logic, tricky edge cases,
52	non-obvious decisions) over trivial boilerplate.
53
54	## Step 4 — Ask one question at a time
55
56	Present a single question, wait for the answer, then evaluate it before moving
57	on. Never dump the whole quiz at once.
58
59	## Step 5 — Evaluate each answer
60
61	- Correct → confirm briefly and move to the next question.
62	- Wrong or incomplete → point to the relevant `file:line` as a hint
63	without revealing the answer, and invite a second attempt.
64	- If the second attempt is still wrong → reveal the correct answer with a
65	short explanation, then continue.
66
67	Track the result of each question (correct / correct-on-retry / missed) to
68	build the final score.
69
70	## Step 6 — Final score and weak areas
71
72	End with:
73
74	- A score, e.g. `Score: 4/6`.
75	- The weak areas: the topics missed, each with a `file:line` pointer to
76	re-read. Keep the tone encouraging, not punitive — the goal is to know what
77	to revisit, not to grade harshly.
78
79	## Optional: automatic suggestion (Stop hook)
80
81	This skill ships an optional, non-blocking Stop hook that suggests running
82	`/handoff` when a session produced a large diff. It never forces the quiz. See
83	[references/install.md](references/install.md) for setup and thresholds.

People who liked this also viewed

6 picks

Type