prompting-skill

A portable skill for building reliable LLM prompts. Drop-in for agentic-stack or any harness that reads SKILL.md frontmatter.

Covers: structure, few-shot, anti-hallucination, output-format locking.

Install into an agentic-stack project

mkdir -p .agent/skills/prompting
cp SKILL.md .agent/skills/prompting/SKILL.md

Then register it by appending to .agent/skills/_index.md:

## prompting
Building reliable LLM prompts — structure, few-shot, anti-hallucination,
output-format locking.
Triggers: "write a prompt", "prompt engineering", "improve this prompt",
"system prompt", "few-shot", "prompt template"

And to .agent/skills/_manifest.jsonl:

{"name":"prompting","version":"2026-04-22","triggers":["write a prompt","prompt engineering","improve this prompt","system prompt","few-shot","prompt template"],"tools":["bash","memory_reflect"],"preconditions":[],"constraints":["do not invent capabilities the model lacks","do not promise deterministic output from a stochastic model","every claim in output must trace to source data in the prompt"],"category":"meta"}

What's inside

The skill distills five load-bearing rules, a production prompt structure, an anti-hallucination checklist, and a worked signature-verification example. See SKILL.md.