writing-agent-instructions

How to write great agent instructions for an agent-native app or template: AGENTS.md, skills, and tool/action descriptions. Use when authoring or reviewing AGENTS.md, writing a SKILL.md, wording action descriptions, or deciding what belongs in instructions vs skills vs memory.

Quality

82%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Passed

No findings from the security scan

Quality

Content

65%

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

A well-organized, actionable creator guide with strong concrete examples, but it runs long for a single SKILL.md, carries some editorial prose and a broken command, and does not practice the progressive disclosure it preaches since no reference files defer its inline depth. Workflow guidance is clear per-topic but lacks a unified validated authoring sequence.

Suggestions

Tighten the editorial sections ('Say each thing once', 'Prefer judgment over rules') into terse imperative bullets, drop the intro framing sentence, and fix the garbled 'pnpm … the repository's guard:agent-chat-context' command so it is executable.

Move the large worked examples (full AGENTS.md template, secrets-hygiene detail, defineAction TS snippet) into references/ files and link to them, keeping SKILL.md a lean overview that practices the progressive disclosure it advocates.

Add an explicit end-to-end authoring checklist with a verification gate — e.g., 'run pnpm guard:workspace-skills and confirm every SKILL.md description has a Use when… clause before considering guidance done' — to give the guide a sequenced workflow with feedback loops.

Dimension	Reasoning	Score
Conciseness	The body is information-dense and mostly product-specific (6000-char cap, tool-search, defineAction), but at ~340 lines it includes editorial prose ('Say each thing once', 'Prefer judgment over rules') and a garbled command ('pnpm the repository's guard:agent-chat-context') that could be tightened — fitting 'mostly efficient but could be tightened' rather than the lean top anchor.	2 / 3
Actionability	Provides concrete, copy-paste-ready examples — a full AGENTS.md template, a defineAction TS snippet, SKILL.md frontmatter samples, and specific commands (pnpm sync:workspace-skills, pnpm guard:workspace-skills) — meeting 'fully executable code/commands; specific examples'.	3 / 3
Workflow Clarity	Per-topic organization is clear and there is one explicit verify step (run guard:workspace-skills before calling guidance done), but there is no end-to-end multi-step authoring workflow with feedback loops; the Do/Don't lists are checklists rather than a sequenced process with validation checkpoints.	2 / 3
Progressive Disclosure	No references/ scripts/ or assets/ bundle files exist, and the skill keeps large depth inline (full AGENTS.md template, secrets-hygiene detail, action-description TS example) rather than deferring it to one-level-deep reference files; cross-links to related skills exist but do not split this skill's own depth.	2 / 3
	Total	9 / 12 Passed

Description

100%

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

A strong description: it states a concrete capability, gives an explicit 'Use when…' trigger covering several natural phrasings, and is tightly scoped to agent-instruction authoring. It is concise and third-person with no over-claims.

Dimension	Reasoning	Score
Specificity	Lists multiple concrete actions and surfaces — 'authoring or reviewing AGENTS.md, writing a SKILL.md, wording action descriptions, or deciding what belongs in instructions vs skills vs memory' — matching the 'lists multiple specific concrete actions' anchor.	3 / 3
Completeness	Clearly answers both 'what' (writing agent instructions for AGENTS.md, skills, and tool/action descriptions) and 'when' via an explicit 'Use when authoring or reviewing…' clause, matching the top anchor.	3 / 3
Trigger Term Quality	Uses natural terms a user authoring agent guidance would actually say — 'AGENTS.md', 'SKILL.md', 'skills', 'action descriptions', 'instructions', 'memory' — with good coverage and no jargon-only phrasing.	3 / 3
Distinctiveness Conflict Risk	Occupies a clear niche (agent-instruction authoring) with distinct triggers (AGENTS.md, SKILL.md, action descriptions) unlikely to fire for unrelated skills; voice is third person so no specificity penalty applies.	3 / 3
	Total	12 / 12 Passed

Validation

81%

Warnings & errors only

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 13 / 16 Passed

Validation for skill structure

Criteria	Description	Result
metadata_version	'metadata.version' is missing	Warning
metadata_field	'metadata' should map string keys to string values	Warning
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning

	Total	13 / 16 Passed

Repository: BuilderIO/agent-native
Path: .agents/skills/writing-agent-instructions/SKILL.md
Commit: e9a2f0e

Reviewed: 4 days ago

Table of Contents

Discovery Implementation Validation

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.