How do I install the Follow Up Handling skill?

Run npx skillmds@latest add ghostlygawd/follow-up-handling in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Follow Up Handling safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Follow Up Handling?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Follow Up Handling free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Follow Up Handling?

GhostlyGawd (@ghostlygawd) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Follow Up Handling by ghostlygawd — Agent Skill

Name: Follow Up Handling
Author: GhostlyGawd

Follow Up Handling

Use at the END of any task that generated deferred work — the "next steps", "you could also", "want me to…", unfixed-nit, or parked-for-later items. Instead of reciting them at the user (which overloads them — their stated correction), capture each via `harness followup add` and end with a single count line. Also use when the user says follow-ups overwhelm them, or asks to defer/park something. Surfacing the list is always a PULL (/followups), never a push.

GhostlyGawd Updated Sep 17, 2026

File contents

Follow-up handling

Follow-ups are real — don't lose them. But reciting them at task-end overloads the user (their explicit correction). Resolve the tension by decoupling capture from surfacing: capture silently, surface only on pull.

The rule

When you finish and notice deferred items — "next steps", "you could also", "want me to…", an unfixed nit, anything parked for later — DO NOT list them in chat. For each, instead:

harness followup add "specific, self-contained, actionable one-liner"

Then end your message with at most ONE line, only if count > 0:

Logged 3 follow-ups (`/followups` to review).

That is the entire surfacing. No "Next steps" section, no "let me know if you want me to…", no re-pitching the deferred work. Zero follow-ups → say nothing.

What is / isn't a follow-up

Yes: deferrable work the user did not ask for now — cleanups, hardening, a related fix, a verification to run later, an optional enhancement.
No — ask directly instead: a blocking question you need answered to finish the CURRENT task. That is not a follow-up; surface it now.
No — just do it: work the user actually requested.

Write them to survive

Each entry must stand alone weeks later: name the file/symbol, the concrete change, the why. "Fix the thing" is useless. "Add templates to the guard BLOCKED-message parenthetical in hooks/guard_enforcement_layer.py" is actionable.

Decay is the point, not a bug

Open follow-ups older than 30 days drop out of the active count automatically (harness followup, TTL). The ledger is a holding pen, not a debt you owe. The user closes items freely (harness followup done <id>) — doing them is optional. Never nag about open follow-ups; the count line and /followups are the only surfacing.

Surfacing is pull-only

The user runs /followups when THEY want the list. You never push it. (A quiet SessionStart count — a number, never the items — may be added later as opt-in.)

provenance: session 56295237, 2026-06-13 — user reported every AI session ends with follow-ups that overload them; routed to capture-silently / surface-on-pull.

---name: follow-up-handlingdescription: Use at the END of any task that generated deferred work — the "next steps", "you could also", "want me to…", unfixed-nit, or parked-for-later items. Instead of reciting them at the user (which overloads them — their stated correction), capture each via `harness followup add` and end with a single count line. Also use when the user says follow-ups overwhelm them, or asks to defer/park something. Surfacing the list is always a PULL (/followups), never a push.---# Follow-up handlingFollow-ups are real — don't lose them. But reciting them at task-end overloads theuser (their explicit correction). Resolve the tension by **decoupling capture fromsurfacing**: capture silently, surface only on pull.## The ruleWhen you finish and notice deferred items — "next steps", "you could also", "wantme to…", an unfixed nit, anything parked for later — DO NOT list them in chat.For each, instead: harness followup add "specific, self-contained, actionable one-liner"Then end your message with at most ONE line, only if count > 0: Logged 3 follow-ups (`/followups` to review).That is the entire surfacing. No "Next steps" section, no "let me know if you wantme to…", no re-pitching the deferred work. Zero follow-ups → say nothing.## What is / isn't a follow-up- **Yes:** deferrable work the user did not ask for now — cleanups, hardening, a related fix, a verification to run later, an optional enhancement.- **No — ask directly instead:** a blocking question you need answered to finish the CURRENT task. That is not a follow-up; surface it now.- **No — just do it:** work the user actually requested.## Write them to surviveEach entry must stand alone weeks later: name the file/symbol, the concrete change,the why. "Fix the thing" is useless. "Add `templates` to the guard BLOCKED-messageparenthetical in hooks/guard_enforcement_layer.py" is actionable.## Decay is the point, not a bugOpen follow-ups older than 30 days drop out of the active count automatically(`harness followup`, TTL). The ledger is a holding pen, not a debt you owe. Theuser closes items freely (`harness followup done <id>`) — doing them is optional. Nevernag about open follow-ups; the count line and `/followups` are the only surfacing.## Surfacing is pull-onlyThe user runs `/followups` when THEY want the list. You never push it. (A quietSessionStart count — a number, never the items — may be added later as opt-in.)provenance: session 56295237, 2026-06-13 — user reported every AI session ends withfollow-ups that overload them; routed to capture-silently / surface-on-pull.

Frequently asked questions