fact-find

majiayu000's avatarfrom majiayu000

Quick lookup of specific facts about Bottlerocket with citations

5stars🔀1forks📁View on GitHub🕐Updated Jan 11, 2026

When & Why to Use This Skill

The Fact Find skill is a high-precision research tool designed to provide instant, cited answers to technical questions about Bottlerocket. It optimizes developer workflows by extracting specific data points—such as configuration defaults, file paths, and versioning—directly from documentation and source code, ensuring all information is verifiable through a structured citation system.

Use Cases

  • Verifying specific configuration values and default settings for Bottlerocket instances to ensure environment consistency.
  • Locating systemd units, targets, and critical file paths for troubleshooting or system automation scripts.
  • Identifying precise version numbers and software dependencies for specific Bottlerocket releases to manage compatibility.
  • Retrieving technical details on disk layouts and partition schemes for infrastructure planning and storage optimization.
namefact-find
descriptionQuick lookup of specific facts about Bottlerocket with citations

Fact Find

Fast, focused answers to specific factual questions about Bottlerocket with proper citations.

Purpose

Quickly find and cite concrete facts about Bottlerocket:

  • Configuration values and defaults
  • Partition schemes and disk layouts
  • Systemd units and targets
  • File paths and locations
  • Version numbers and dependencies

When to Use

  • Need a specific fact, not an explanation
  • Question has a concrete, definitive answer
  • Looking for "what is" or "where is" information

For broader questions about architecture or design, use deep-research instead.

Roles

Role Reads Does
Orchestrator (you) SKILL.md Creates workspace, spawns subagent, returns answer
Subagent phases/FACT-FIND.md Searches, reads files, writes cited answer

⚠️ You do NOT read the phase file — pass it to the subagent via context_files. The subagent handles all search and file reading, keeping that context out of yours.

Orchestrator Instructions

workspace = "planning/<question-slug>"
mkdir workspace
write workspace/question.txt with the user's question

result = spawn(
    prompt = "Answer the factual question in the workspace.",
    context_files = ["skills/fact-find/phases/FACT-FIND.md"],
    context_data = {"workspace": workspace},
    allow_tools = True
)

read workspace/ANSWER.md
present to user

Inputs

  • User's factual question about Bottlerocket

Outputs

  • workspace/ANSWER.md: Concise answer with inline citations, sources section, and Research Quality Indicator

Citation Format

The final answer uses this format:

<Answer text with inline citations <sup>[1]</sup>.>

## Sources

<sup>[1]</sup> [`path/to/file.md`](../path/to/file.md)
- What this source provided

---

✅ **Answered from documentation** | ⚠️ **Answered from source code** | 🔍 **Partial documentation**

Validation

A good fact-find response:

  • ✓ Directly answers the specific question
  • ✓ Concise (2-4 sentences typically)
  • ✓ Superscript citations inline
  • ✓ Sources section with numbered references
  • ✓ Research Quality Indicator at end
  • ✓ No unnecessary context or explanation
fact-find – AI Agent Skills | Claude Skills