agent-output

Compresses agent-to-agent reports into a structured envelope with state, evidence, and next actions.

14|1|Updated Jun 14, 2026
One-click install
npx skills add https://github.com/allemaar/open-skills --skill agent-output-allemaar
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: agent-output
Source: https://github.com/allemaar/open-skills/tree/main/skills/agent-output
Command: npx skills add https://github.com/allemaar/open-skills --skill agent-output-allemaar

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Subagent and worker reports are often verbose, burying the verdict, evidence, and required action under chronology and filler. This Skill enforces a compact communication contract so receiving agents can decide, verify, continue, or stop without decoding long prose. ## Core Features & Use Cases - Structured report envelope: Standard fields (STATE, RESULT, EVIDENCE, CHANGED, GAPS, NEXT) for worker reports, plus FINDING/IMPACT/FIX variants for review findings and a minimal two-line form for live status updates. - Compression rules: Lead with the result, name exact artifacts, merge repeated findings by root cause, and send only one recommendation, with no greetings or recaps. - Fidelity floor: Brevity never removes authority limits, stop conditions, destructive-action gates, exact numbers, uncertainty, or failed checks. - Use Case: When dispatching a subagent to investigate a bug, include this return contract in the worker brief so the report comes back as a terse, decision-ready envelope instead of a narrative log. ## Quick Start Ask the agent to report its completed task using the agent-output envelope with STATE, RESULT, EVIDENCE, and NEXT fields.

Frequently Asked Questions about agent-output

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I make subagent reports shorter and more useful?

Use a fixed report envelope with STATE, RESULT, EVIDENCE, CHANGED, GAPS, and NEXT fields, omitting empty ones. Lead with the result, name exact artifacts instead of pasting logs, and include only one next action per report.

What format should AI agents use for status updates and handoffs?

For full reports use the six-field envelope; for findings use FINDING, IMPACT, EVIDENCE, FIX, and GAPS; for live updates send only STATE plus the changed fact. Match the envelope size to the message so two lines suffice when they carry the point.

Does compressing agent reports lose critical safety information?

No. The fidelity floor requires that brevity never removes authority limits, stop conditions, destructive-action gates, exact numbers, scope, uncertainty, attribution, failed checks, or untested areas. A short report hiding a reversing caveat is invalid.

What is the difference between agent-output and agent-mailbox?

agent-output defines the compact message format and compression rules for agent reports. agent-mailbox handles durable message transport and delivery between agents, so the two skills solve formatting versus delivery.

When should I not use the full report envelope?

Skip the full envelope for short live updates where only STATE plus the changed fact carries the message. Sending all six fields for a two-line update adds noise without improving the receiver's decision.