How to Use Claude Skills to Turn an On-Site Incident Note Into a Clear Incident Report

Warning: Incident notes can contain legal, HR, health, or insurance-sensitive information. Use only the Claude setup your business has approved and keep the workflow limited to the exact files needed.

If Claude Skills is new to you

Start with the Claude Skills explainer page in the site's Tools menu before you run this workflow. The short version is that Skills are reusable instruction bundles Claude can load for repeatable work. The public skills repository also makes good reference material when you want to copy the pattern into a private skill for your own team. For curated examples, see the Claude Skills tool hub and the Anthropic Workflows series.

Problem and who this is for

Something happened on site. The first notes are rushed and incomplete. You may have a handwritten note, a screenshot of a witness text, and a short verbal summary from whoever was present. You need one clear incident report that is factual, neutral, and useful later.

This workflow is for local business owners and operators who need an internal incident report without turning messy source material into a liability.

Prerequisites

• Claude Skills access.
• At least one source note or image.
• Any related witness text screenshots or manager notes.
• A rule that uncertain details stay uncertain.

How to gather the source material

1. Photograph handwritten notes in good light.
2. Save screenshots with descriptive names like witness-text-1.png.
3. Create one text file called incident-basics.txt with the date, location, people involved if known, and whether external follow-up already happened.
4. If a source includes speculation or emotion, keep it, but label it as a reported statement during review.
5. Do not combine unrelated incidents in the same run.

Numbered workflow steps

1) Start with fact extraction, not polishing

The internal-comms skill pattern is useful here because it already covers incident reports. The safe path is to extract facts first, then write the final report from the verified extraction.

2) Run a facts-only extraction pass

{
  "task": "Extract incident facts with no guessing",
  "input": {
    "incident_sources": "ATTACHED NOTES, IMAGES, SCREENSHOTS",
    "incident_basics": "PASTE incident-basics.txt"
  },
  "rules": [
    "Do not add causes, motives, or blame.",
    "Preserve direct quotes exactly.",
    "Separate facts from reported statements.",
    "Mark missing or unclear details [UNCLEAR].",
    "Keep chronology explicit."
  ],
  "output": {
    "factual_timeline": "Chronological list of what is documented",
    "people_and_roles": "Known people and roles only",
    "direct_quotes": "Verbatim quotes from texts or notes",
    "reported_statements": "Claims that are not independently verified",
    "uncertainties": "What still needs confirmation"
  }
}

3) Verify the high-risk fields manually

Check names, dates, times, locations, and direct quotes. If you know a correction, add it as a separate verified note. Do not silently overwrite the original source wording.

4) Generate the incident report

{
  "task": "Write a clear internal incident report from verified incident facts",
  "input": {
    "verified_incident_input": "PASTE STEP 2 OUTPUT AFTER REVIEW",
    "tone": "neutral, factual, concise"
  },
  "rules": [
    "Keep causality unknown unless the source explicitly states it.",
    "Do not assign fault.",
    "Keep reported statements separate from confirmed facts.",
    "Use placeholders for missing fields instead of guessing."
  ],
  "output_format": {
    "type": "incident_report",
    "sections": [
      "Summary",
      "Known facts",
      "Timeline",
      "Reported statements",
      "Open questions",
      "Follow-up owner or next check"
    ]
  }
}

Tool-specific instructions

Claude Skills

• Read the Claude Skills explainer page in the site's Tools menu if you need a quick overview first.
• The internal-comms skill pattern is especially relevant because its documented use cases include incident reports and internal updates.
• If your business wants a fixed incident form later, create a private custom skill with your own field order and label rules.

Quality checks

1. The final report preserves the difference between facts, quotes, and unverified statements.
2. Names, dates, times, and locations match the verified sources.
3. Missing details remain visible instead of being guessed.
4. The report would still make sense to someone reviewing it later with no extra context.

Common failure modes and fixes

Claude turns a witness claim into a fact

Fix: rerun Step 2 and explicitly require a separate reported_statements section.

The report sounds accusatory

Fix: tell Claude to remove causal or blame language and keep only verifiable wording.

Too many details are still unclear

Fix: generate the report anyway with clear uncertainty markers, then append a short follow-up checklist for the manager to confirm.

Sources Checked

• Anthropic Claude Help Center: What are Skills? (accessed 2026-03-23) https://support.claude.com/en/articles/12512176-what-are-skills
• Anthropic Claude Help Center: Use Skills in Claude (accessed 2026-03-23) https://support.claude.com/en/articles/12512180-use-skills-in-claude
• Anthropic Claude Help Center: How to create custom Skills (accessed 2026-03-23) https://support.claude.com/en/articles/12512198-how-to-create-custom-skills
• GitHub: elliepetalmedia/skills README (accessed 2026-03-23) https://github.com/elliepetalmedia/skills
• GitHub: internal-comms skill in elliepetalmedia/skills (accessed 2026-03-23) https://github.com/elliepetalmedia/skills/tree/main/skills/internal-comms

Quarterly Refresh Flag

Review on 2026-06-21 to confirm any updates to Claude Skills, Cowork availability, mobile dispatch requirements, project behavior, and supported setup paths.