Skip to content
Agentmart
Submit a tool

writing-research-skills

Design, review, and maintain predictable Research OS skills.

Skill by yhlai0911

IndexedMIT
Write a review

Chalk Labs

Need help deploying writing-research-skills?

Chalk Labs sets up agents like this inside real teams: connected to your tools, tested on your data, with a human in the loop where it matters.

Book a call

Install

Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install skill/yhlai0911-writing-research-skills

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

Design, review, and maintain predictable Research OS skills.

Use cases: design, writing, research

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.

Tools
No tools found
Secrets
None asked for
Install scripts
None found

Network hosts it talks to 1

  • codeload.github.com · found in install command

Files it touches

  • ~/.claude/skills/writing-research-skills

What's inside

Skill name
writing-research-skills
Description
Design, review, and maintain predictable Research OS skills.
License
MIT
Instructions
78 lines in SKILL.md · view on GitHub

Files (3)

  • SKILL.md
  • openai.yaml · agents/
  • design-vocabulary.md · references/

SKILL.md (first lines)

# Writing Research Skills

Optimize for **predictability**: the skill should drive the same process across runs while allowing the research output to vary with evidence.

Read [design-vocabulary.md](references/design-vocabulary.md) when diagnosing structure, invocation, disclosure, or drift.

## Process

### 1. Ground the behavior

Collect at least three concrete user prompts that should trigger the skill and two near-misses that should not. For each true case, state the artifact, decision, or observable state the skill must produce.

**Complete when:** trigger boundaries and expected outcomes are concrete enough to forward-test without hidden context.

### 2. Choose the invocation boundary

Use model invocation only when the agent should autonomously reach the skill or another skill needs its discipline. Use user invocation for explicit orchestration, state changes, or long flows. Keep user-only skills behind `policy.allow_implicit_invocation: false`.

Give each model-invoked skill one distinct leading concept that also appears in its trigger description. Prefer one strong primitive over several synonymous skills.

**Complete when:** the invocation choice, leading concept, and unique responsibility differ from every catalogued skill.

### 3. Design the contract

Define:

- inputs and factual discovery policy;
- decisions reserved for the researcher;
- canonical reads and writes; for workspace skills, read only `.research/config.yaml` first, resolve its adapters, and never probe a baseline path before resolution;
- one current-artifact registry in the resolved context, with durable pointers and `complete`, `partial`, or `blocked` outcomes;
- ordered steps or flat reference rules;
- a checkable, exhaustive completion criterion for every step;
- blocked, partial, and stop branches;
- external side effects and their authorization gate.

State positive target behavior. Use prohibitions only for hard safety boundaries and pair them with the intended action.

**Complete when:** an agent can distinguish complete, partial, and blocked from artifacts and receipts rather than prose confidence.

### 4. Allocate context

Keep essential actions in `SKILL.md`. Move branch-specific detail to one-level `references/`, deterministic operations to `scripts/`, and copyable output material to `assets/`. Keep each meaning in one authoritative place.

Split only when a new skill deserves autonomous invocation or when a real context boundary must hide later steps to prevent premature completion.

**Complete when:** every line changes behavior, every pointer says when to load its target, and no reference is duplicated.

### 5. Initialize and implement

Add the skill to `skills/catalog.yaml`, then use Codex’s official `skill-creator` initializer. Replace all generated placeholders. Keep frontmatter to `name` and `description`; regenerate `agents/openai.yaml` from the catalog.

**Complete when:** no initializer marker, placeholder, stale name, or unowned shared rule remains.

### 6. Prove predictability

Run:

```bash
uv run python scripts/sync_catalog.py --check
uv run python scripts/validate_repository.py
uv run pytest
```

Forward-test realistic prompts in fresh agents. Pass the skill and raw task, not the intended answer. Include one success case, one near-miss trigger, and one blocked case. Repair the smallest source-of-truth rule that explains a failure, then rerun.

**Complete when:** structural checks pass and forward tests follow the intended process without leaked conclusions.

## Failure modes

- **Premature completion:** sharpen the current completion criterion before splitting the sequence.
- **Duplication:** choose one authority and replace copies with pointers.
- **Sediment:** remove stale instructions instead of adding exceptions.
- **Sprawl:** disclose branch-specific reference or split at a real boundary.
- **No-op:** delete instructions that do not change default model behavior.
- **Trigger collision:** merge synonyms or give each skill a genuinely distinct leading concept.
- **Truth drift:** generate inventories and metadata from the catalog; do not maintain parallel hand-written lists.

From yhlai0911/research-os, MIT.

Signals

Stars
0
Forks
0
Open issues
0
Last commit (this folder)
Jul 26, 2026 (3 months ago)
Last release
Oct 19, 2026 (29 days ago)

From GitHub (fetched Oct 24, 2026).

Details

License
MIT
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

Done for you

Rather not set it up yourself?

Chalk Labs scopes, builds and runs agents for your team: the task audit, the eval harness, the integrations and the weekly ops report. Fixed scope, priced before we start.

See how we build AI agents →Read the case studies

Chalk Labs is a two-founder AI and growth studio. We build AI agents, CRMs and MVPs for Web3 and AI companies, and have shipped 50+ product and token launches.

launches shipped
50+
launches shipped
media placements
500+
media placements
founders on every account
2
founders on every account

Press we've placed: Forbes · CoinDesk · Decrypt · Cointelegraph · The Block

Rahil Jain, Co-founder, Chalk Labs. Ex-Goldman Sachs engineer. You talk to the person who builds it.

Get writing-research-skills running in your business

Tell us what you want to automate. A founder replies within one working day with a plan, not a pitch deck.

Goes to the Chalk Labs founders, not the newsletter. Privacy

Need help deploying writing-research-skills?

CallGet help
Cite this page

writing-research-skills. Agent Mart. https://agentmart.dev/skills/yhlai0911-writing-research-skills (last updated 2026-10-24)