agentleFS
agiea-brainagiea-brain/agency-skills/skills/hunter-harris/kondo-owed-reply-sweep/SKILL.md

kondo-owed-reply-sweep

Skill1 starsChanged 17 days agoMITView on GitHub

Use this skill at the start or end of a day of LinkedIn prospecting, when you need to know which DMs are waiting on you. Produces a short list of people whose message is last, for you to decide which to answer. Uses Kondo's MCP server. "Who replied?", "check my DMs", "what LinkedIn replies do I owe".

The file

We split this skill into the rules it gives an agent

8 rules. About 0 are common, 6 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈6

    Distinctive · estimate

    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:

    Call list_inbox with the "needs my reply" view and an after cutoff 14 days back, computed in UTC epoch milliseconds.

    line 20
  2. Distinctive:

    If a chat comes back in pages, merge the pages into one thread before judging it.

    line 24
  3. Distinctive:

    If you can check your email, texts or call log, suppress threads you already answered there after their last message.

    line 26
  4. Distinctive:

    MUST stay read-only: never send, draft into Kondo, label or archive a chat from this sweep.

    line 40
  5. Distinctive:

    MUST name people, never show Kondo's internal identifiers.

    line 41
  6. Distinctive:

    MUST let the human decide who needs a reply.

    line 42
  7. Distinctive:

    NEVER record anything when Kondo is down or returns an error.

    line 43
  8. Very specific:

    NEVER fire load_chat calls in parallel.

    line 44

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “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

Applies when you prospect on LinkedIn and replies get lost in the inbox. Produces a list of the one-to-one threads where the other person spoke last, by name, for you to decide.

Needs Kondo connected to your agent through its MCP server. The sweep only reads, covers the last 14 days, and sees only LinkedIn.

The play

  1. Call list_inbox with the "needs my reply" view and an after cutoff 14 days back, computed in UTC epoch milliseconds. Use a fixed helper or a one-line script for the cutoff, never the agent's head. Save the raw response to a file.
  2. An error or missing response means the sweep failed: say so and stop, never log it as done. A valid empty inbox means nothing is owed.
  3. Drop group chats; they are never a reply owed. Drop threads you already marked "no reply needed", unless the person has written since.
  4. For each remaining chat, call read_chat. If it comes back empty, call load_chat (also a read), spacing calls 3 seconds apart. If rate limited, wait as told and retry once.
  5. If a chat comes back in pages, merge the pages into one thread before judging it.
  6. Sort by time and look at the last real message. It is owed when the last message is theirs. An attachment with no text still counts as a message.
  7. If you can check your email, texts or call log, suppress threads you already answered there after their last message. If you can't, keep them on the list.
  8. List the owed people by name and ask once which to answer. For each one you say needs no reply, store the thread and the time of their last message, so it comes back only when they write again.
  9. Close with one line: owed, answered, not owed.

The failure log behind steps 1 to 7 is in references/sweep-gotchas.md. Open it when a count looks wrong.

What good looks like

  • First check: the sweep really ran. An error string logged as "done" means a day of replies nobody saw.
  • The mediocre version trusts Kondo's "needs reply" view as-is. It flags group chats, splits one paged chat into several, and asks every morning about people you answered by text.
  • The output is right when every name on the list has their message last, nobody you answered elsewhere is on it, and the list is short enough to act on before the day's messaging starts.

Rules

  • MUST stay read-only: never send, draft into Kondo, label or archive a chat from this sweep.
  • MUST name people, never show Kondo's internal identifiers.
  • MUST let the human decide who needs a reply; record their words when they say "no reply needed".
  • NEVER record anything when Kondo is down or returns an error.
  • NEVER fire load_chat calls in parallel.

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.