Skip to content
Back to skills

Task Prompts

ASecurity

Write grounded task, delegation, and continuation prompts for coding agents.

  • 10 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added September 23, 2026
ai-agentsgogitdocumentation

Works with

  • cli

Security analysis

A100/100

Pro scans all 2 files and shows the line behind each finding

Scanned October 6, 2026

npx -y skills add fmind/dot --skill task-prompts --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Task Prompts?

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

Security grade badge for Task Prompts
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/fmind-task-prompts/badge)](https://www.skillsdirectory.com/skills/fmind-task-prompts)

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

Download with Pro
SKILL.md
---
name: task-prompts
description: "Write grounded task, delegation, and continuation prompts for coding agents."
license: MIT
metadata:
  kind: task
  author: Médéric HURIER (Fmind)
  source: github.com/fmind/dot/tree/main/skills/task-prompts
  created: "2026-09-05"
  updated: "2026-10-05"
---

# Task and Continuation Prompts

Prepare instructions the receiving agent can use without this conversation. [prompt-design](../prompt-design/SKILL.md) owns prompts embedded in applications; native resume or compaction is preferable when it already preserves the needed state.

## Workflow

1. **Choose the mode**: fresh task, or continuation of work already underway; resolve the requested outcome, latest corrections, scope, and acceptance criteria.
1. **Ground the prompt**: inspect relevant repository instructions, source, tests, and installed dependencies; verify paths and commands and label assumptions.
1. **Record continuation state**: check `git status --short --branch`, record completed proof, the exact stopping point, outstanding work, failed approaches, and reasons for material decisions. Before `/clear` or a harness switch, use [handoff](../handoff/SKILL.md), which also copies the result to the clipboard.
1. **Draft** from [prompt-template.md](references/prompt-template.md); omit empty sections and add ordered slices only when execution needs them.
1. **Write and check**: resolve the root with `git rev-parse --show-toplevel`; save `.agents/prompts/<file>.md`, defaulting to `<YYYY-MM-DD>-<slug>.md`, or `~/.agents/prompts/` outside a repository. Verify the receiver can act without hidden context, then report the path.

## Gotchas

- **Authority travels with the task**: distinguish authorized actions from proposed work; writing the prompt does not grant additional permission.
- **Keep history useful, not copied**: preserve failures and proof gaps; cite code paths and lines instead of copying source or the shared persona.
- **Keep the prompt inbox untracked**: keep `.agents/prompts/` gitignored unless the user wants it tracked; keep a small continuation short.

## Documentation

- Companion skills: [research-brief](../implementation-plan/references/research-brief.md) (decision evidence), [implementation-plan](../implementation-plan/SKILL.md) (ordered slices), [agent-project](../agent-project/SKILL.md) (host layout), [handoff](../handoff/SKILL.md) (save and copy before `/clear`).

Files in this skill

  • SKILL.md2.1 KB
  • references/prompt-template.md1.3 KB

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…