Katok reply reuse

Skill mrcha033/skills/plugins/katok-reply-reuse/skills/katok-reply-reuse

MrCha Skills — portable Agent Skills for ChatGPT, Codex, and compatible agents

Install

npx -y skills add mrcha033/skills --skill katok-reply-reuse

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

3 things to look at

16 days oldThe repository was created 16 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Find and reuse the user's own past KakaoTalk replies for a new incoming message, with provenance and only explicit placeholder substitution. Use when the user asks what to reply, wants their real KakaoTalk tone, dislikes AI-sounding rewrites, or wants a past response to a similar situation retrieved from a macOS KakaoTalk archive.

SKILL.md

2.1 KB, as published. Nobody here has run it

KakaoTalk Reply Reuse

Retrieve first. Do not generate a polished AI reply and then try to humanize it.

Workflow

Run katok doctor --json. Report archive/index readiness without opening message content.
Obtain the incoming message and enough relationship/context terms to search.
Search the authorized archive with the KakaoTalk archive skill. Retrieve explicit matching chunks only when the user has asked to use their messages for this reply.
Keep only records that metadata identifies as authored by the user. If authorship cannot be established, do not use the text as the user's exemplar.
Build an authorized JSONL candidate file following references/reuse-contract.md.

Rank it:

python3 "$SKILL_DIR/scripts/rank_replies.py" \
  --input candidates.jsonl \
  --query "새로 받은 메시지" \
  --limit 3

Prefer the highest-confidence exact past reply. Substitute only placeholders explicitly provided with --slot key=value.
Show the source date/chat label, similarity score, and what changed. Never send automatically.

Abstain

If no strong exemplar exists, say so and ask for either a rough draft or a known similar conversation. Do not silently fall back to generic generation. Use a generated fallback only when the user explicitly requests one, and label it as generated rather than retrieved.

Privacy

Read the smallest number of chunks needed.
Do not expose unrelated participants or messages.
Do not access the underlying KakaoTalk database directly.
Do not write private messages into the skill repository, tests, logs, or examples.
Use only synthetic data in scripts/self_test.py.

Run python3 "$SKILL_DIR/scripts/rank_replies.py" --self-test after modifying ranking.