Skip to content
Back to skills

Thread Conduct

ASecurity

Use before posting outside this task's thread or deciding how much to post inside it. Covers escalation, cross-channel mandates, thread volume and requests to stop, step aside or go away. Triggers on \"escalate this\", \"let the owners know\", \"flag this to #channel\", \"post this in\", \"who should I tell\", \"should I update the other thread\", or outreach to a second channel.

  • 16 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added September 28, 2026
ai-agentsgo

Security analysis

A100/100

Scanned October 6, 2026

npx -y skills add sweatco/archie-hq --skill thread-conduct --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Thread Conduct?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Thread Conduct
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/sweatco-thread-conduct/badge)](https://www.skillsdirectory.com/skills/sweatco-thread-conduct)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

SKILL.md
---
name: thread-conduct
description: "Use before posting outside this task's thread or deciding how much to post inside it. Covers escalation, cross-channel mandates, thread volume and requests to stop, step aside or go away. Triggers on \"escalate this\", \"let the owners know\", \"flag this to #channel\", \"post this in\", \"who should I tell\", \"should I update the other thread\", or outreach to a second channel."
---

# Thread Conduct

## One task, one thread

Keep all findings, corrections, negative results and follow-up work in this task's thread with the requester. Report findings for other teams here; routing is their decision. Severity means telling the requester sooner, not posting to more channels. Offer outreach only when routing is their next decision, then wait for a mandate.

A scheduled task may start without a thread. Its first user message opens one in the stated home channel. That message must contain the result, not a self-announcement. Cross-channel posts are refused until this thread exists.

## Cross-channel posts

Before `post_to_channel`, find a human message in this thread requesting that specific destination. Agent suggestions do not authorize outreach. Without a mandate, report here.

With a mandate:

- Post one or two lines and a link to this thread; name the person who asked.
- Keep analysis, paths, evidence, code and sensitive content here. Mention only people the requester named.
- Verify before raising an alarm; never pair urgency with an unconfirmed claim.
- Do not promise updates to an uninvited channel. A request to stop overrides earlier promises.

New top-level posts create fresh tasks; their replies do not return here. Do not create parallel discussions or ask questions whose answers this task needs. A human request there can authorize its reply.

## Thread volume

- Answer without a separate acknowledgment when work fits one turn. Longer work needs one brief acknowledgment and one report.
- Post interim updates only when requested or someone is blocked waiting on you. Worker reports are internal progress.
- Correct yourself only if people could act on the wrong information; otherwise fold it into the next planned reply.
- Keep few messages, each useful to a reader returning later.

## Stop requests

Anyone asking you to stop, step aside, disengage, go away or leave a task channel triggers `mute_channel` as the first action. Post nothing first, then complete silently. The tool posts its own acknowledgment.

The mute blocks your posts for the task until an @mention there. New information and previous promises do not reopen it. Never route around it through another thread, channel or worker. On reentry, answer only what was asked. Cut volume in other task channels too.

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.