writing-agent-instructions
Use when writing or tightening standing instructions an agent follows across many cases,
Install / Use
npx skills add synapseradio/ai-skills --skill writing-agent-instructionsInstalls into whichever agent you are using.
SKILL.md
Installable skill definition
Quality Score
Category
Content & MediaSupported Platforms
Tags
Our assessment of writing-agent-instructions
writing-agent-instructions scores 59/100 on our quality scale, 1104th of 1,143 Content & Media skills we index.
Its SKILL.md is 4.3 KB long, well organised into 12 sections with 1 code example: a solid amount of guidance for an agent.
It has no GitHub stars yet, so there is no community track record; judge it on its content.
Maintenance, license and trust
- We could not determine when the repository was last updated.
- No license is declared. By default that means all rights are reserved: you can read it, but reusing or redistributing it is not clearly permitted. Ask the author before building on it commercially.
- Its trust signals score 68/100, with 3 cautions from licensing, adoption, age or documentation. These come from repository metadata, not a code audit — read the skill file before letting an agent act on it.
writing-agent-instructions compared with similar skills
All 4 of these similar skills score higher than writing-agent-instructions; compare them before choosing.
| Skill | Score | Stars | Updated | Format |
|---|---|---|---|---|
| writing-agent-instructions (this skill)by synapseradio | 59 | 0 | — | SKILL.md |
| siyuanby siyuan-note | 100 | 47.3k | today | MCP Server |
| algorithmic-artby anthropics | 100 | 177.9k | 58d ago | SKILL.md |
| pptxby anthropics | 100 | 177.9k | 58d ago | SKILL.md |
| prospectingby coreyhaines31 | 100 | 60.1k | 8d ago | SKILL.md |
Frequently asked questions
- How do I install writing-agent-instructions?
- Run
npx skills add synapseradio/ai-skills --skill writing-agent-instructions. The install tabs above show the steps for each supported agent. - Which AI agents does writing-agent-instructions work with?
- It is written for Universal, as a SKILL.md file. Other agents that read the same format can often use it too.
- Is writing-agent-instructions safe to use?
- It declares no license and scores 68/100 on trust signals. Skills are instructions an agent will follow, so read the file before installing it and do not approve commands you do not understand.
- Is writing-agent-instructions still maintained?
- We could not determine when the repository was last updated.
Skill content
View source on GitHubname: writing-agent-instructions description: Use when writing or tightening standing instructions an agent follows across many cases, such as rules or system guidance. Shapes them so every reader applies them the same way, including to cases beyond the ones the writer listed.
Writing agent instructions
How an instruction works
An instruction is read by a model that meets cases the writer never saw. Hold the instruction as a few relations the reader evaluates on the case in front of it, and the acts it yields as the smallest set the text derives. A list of cases derives only its rows, and the reader handles the rest by analogy, which varies from reader to reader. A generator derives the rows and every case between them, the same way in every reader, because each reader runs the same test on the same visible features.
The writer's work is to name the set, name the feature that decides the act, say how the arms combine, and then trust the reader's expansion while proving it: expand the text by hand on the cases in hand and on cases nobody listed, and show every act the text derives that nobody asserted, before it lands. A derived act nobody wrote goes to the approval list before it enters the file.
A generator has four parts. The domain names the set of cases it covers, with a test a reader can run on any candidate. The mapping names the feature of a case that decides the act, and the act for each value. The expansion rule says how the reader turns the mapping into an act, "take the first arm that fits" for one. The anchor is one example at most, placed where the mapping could be read two ways.
Route first
Decide the route from the state of the world. Read a route's reference when you enter it, not before.
- No instruction exists yet, only a need, an incident, or a behavior to capture → Write. Read references/write.md on entering. Write ends by entering Refine with its draft.
- An instruction exists, or a draft was just written, and it should say the most in the least space, or the question is whether it works or what a rewrite lost → Refine. Read references/refine.md on entering.
- Neither, as when the file is short and the ask is a wording change → say so and make the change without the procedure.
When you cannot tell which route fits, ask one question rather than guessing. Where nobody can answer, return that question with what each route would produce.
Where the run can open or write no file, as when the instruction arrives pasted into the conversation, take the pasted text as the file: do each search by reading the text whole, and return the output in the reply for whoever holds the file to place.
Both routes end in the same artifacts, the four parts and the checks in references/checks.md, and differ only in where the cases come from. Close a route only when its output exists in the shape at the end of this file.
Wording
Word each line as the act to take. Where a draft line says "never X", write the act to take in its place, a gate ("X only on Y"), or an omission ("Keep X out of Y"), since a reader acts on what a line names and a ban names only what to avoid. Name a feature in the negative only where the feature is already defined on the page. Where a line states a fact, write it as a stance the reader holds, "treat", "count", or "hold". Define each coined term where it first lands. Keep intensifiers out, since a reader weighs every line the same.
Output
## Generator
[the domain, mapping, expansion rule, anchor, and glosses, as they will appear in the file]
## Residue
[each kept line, with its form; "none" where empty]
## Accounting
[the trace table, complete]
## Disagreements
[where a rewrite was handed in: each trace row where your expansion and the rewrite's own accounting differ; leave the heading out otherwise]
## For approval
[each edge row with no line already deciding it, as a yielded act nobody wrote]
## Dropped
[each original line, pruned line, or collected case that neither the generator nor the residue carries, with its meaning named; "none" where nothing drops]
## Size
[characters before and after, where an original exists]
Drop nothing silently. Every original line and every collected case appears in the accounting, the residue, or the dropped list.
Related Skills
siyuan
47.3kAn open-source, privacy-first, self-hosted knowledge workspace where humans and AI agents work together 开源、隐私优先、自托管的知识工作空间,让人与智能体在此协作
algorithmic-art
177.9kCreating algorithmic art using p5.js with seeded randomness and interactive parameter exploration. Use this when users request creating art using code, generative art, algorithmic art, flow fields, or particle systems.
pptx
177.9kUse this skill any time a .pptx or .potx file is involved in any way — as input, output, or both. This includes: creating slide decks, pitch decks, or presentations; reading, parsing, or extracting text from any .pptx or .potx file (even if the extracted content will be used elsewhere, like in an em…
prospecting
60.1kWhen the user wants to find and qualify prospects for outbound, or asks who to target first, which signal to prioritize, or where to spot buyers with the problem: buying signals, enrichment, verification, account research.
Trust signals
From repository metadata: license, adoption, age and documentation. Not a code audit — see the Safety scan above for what the skill file itself contains.
