Purpose
Turn a raw quality flag or error observation into a written feedback message that is specific, fair, and easy to act on — one that corrects the behavior without demoralizing the recipient. Solves the problem of feedback that is either too soft to change anything or too blunt to land well.
When to Use
Invoke this skill when the user asks to:
- Write feedback to an agent or team member about a specific error or mistake
- Turn a QC finding, audit flag, or scorecard result into a coaching message
- Deliver correction on a missed standard, SOP deviation, or quality issue
- Draft a written note before a 1:1 or coaching conversation
- Reframe blunt or reactive feedback into a constructive version
- Give recurring-issue feedback where a prior warning already exists
Inputs Required
Before starting, collect:
- Recipient name: The person receiving the feedback (first name)
- Error type: What went wrong — the specific mistake, deviation, or missed standard
- Context: When/where it happened, what task or interaction, any ticket/call/order reference
- Correct approach: What the person should have done instead (if unknown, the skill infers a reasonable standard and flags the assumption)
- Severity: Minor / Moderate / Serious — or describe the impact and let the skill classify
- History: Is this a first occurrence, a repeat, or after a prior warning?
- Relationship/tone target: Peer, direct report, new hire, high performer, or under-performer
If severity or history is missing, treat it as a first-occurrence moderate issue and note the assumption.
Instructions
Step 1: Classify Severity and Set the Frame
Map the error to a severity tier — this drives the message's directness and length.
- Minor (cosmetic, low-impact, easily corrected): Light-touch, encouraging, brief. Lead with the fix, not the failure.
- Moderate (affects quality, customer experience, or a metric; repeatable): Balanced structure, clear expectation, explicit next step.
- Serious (policy breach, customer harm, compliance risk, or a repeat after warning): Direct and unambiguous. State the standard, the consequence path if relevant, and the required corrective action.
Adjust for history: a repeat issue moves the tone one notch firmer and explicitly references the prior instance.
Step 2: Build the Message Using the SBI-N Framework
Structure the core of every message around four beats:
- Situation — Name the specific context: task, ticket, time, or interaction. No vague "recently" or "sometimes."
- Behavior — Describe exactly what was done, observably. Fact, not interpretation. Avoid "you always" or character judgments.
- Impact — Explain the concrete downstream effect: on the customer, the metric, the team, or the record. This is what makes feedback land.
- Next step — State the correct approach and one clear, doable action going forward. End on a forward path, not the failure.
Step 3: Calibrate the Open and Close
- Open: Warm but direct. Acknowledge the person by name. For high performers, lead with a genuine strength before the correction. Skip filler praise for repeat serious issues.
- Close: Confidence + door open. Reaffirm that the fix is achievable and offer support ("Let me know if you want to walk through it"). For serious issues, replace the soft close with a clear checkpoint ("Let's confirm this is corrected on the next review").
Step 4: Apply the Fairness and Tone Filters
Before finalizing, run the message through these filters:
- Specific, not global: Corrects one behavior, not the person's competence
- Balanced, not padded: No "compliment sandwich" that buries the point — the correction stays clear
- Actionable, not abstract: The recipient knows exactly what to do differently tomorrow
- Owned language: "The standard is…" / "What we need is…" rather than personal blame
Step 5: Produce the Deliverable
Output the ready-to-send feedback message, followed by a short delivery note (channel and timing recommendation) so the operator knows how to send it, not just what to send.
Output Format
- Structure: One ready-to-paste feedback message + a brief delivery note
- Length: 120–220 words for the message (shorter for Minor, longer for Serious); delivery note 2–3 sentences
- Format: Prose message written directly to the recipient, using their first name; delivery note as a short labeled block
- Sections: Feedback Message → Delivery Note (recommended channel, timing, and whether a live conversation should accompany it)
- Tone: Coaching, respectful, direct — a good team lead who wants the person to succeed. Never sarcastic, passive-aggressive, or falsely cheerful.
Quality Checklist
Before delivering output, verify:
- [ ] The message names a specific situation — not "recently" or "a few times"
- [ ] Behavior is described observably, with no character judgments ("you always," "you're careless")
- [ ] The impact of the error is stated concretely (customer, metric, team, or record)
- [ ] Exactly one clear next step or correct approach is given
- [ ] Tone and directness match the severity tier and history
- [ ] The message uses the recipient's name and closes on a forward path or checkpoint
Notes
- The most common failure in written feedback is skipping Impact — without it, the recipient hears an opinion, not a reason to change
- For repeat serious issues, keep the message factual and documentable; it may become part of a performance record
- Written feedback works best for minor/moderate issues; recommend pairing serious feedback with a live conversation, and use the message as the shared reference
- Avoid the compliment sandwich for anything above Minor — it dilutes the correction and recipients learn to wait for the "but"
How to use this skill
This file is a Claude Code skill in Anthropic's standard format — markdown with YAML frontmatter at the top.
Installing in Claude Code:
~/.claude/skills/<folder-name>/SKILL.md— auto-loads in every session anywhere on your machine (recommended for skills you'll use across multiple projects)<your-project>/.claude/skills/<folder-name>/SKILL.md— auto-loads only when working in that project (recommended for project-specific skills)
Restart Claude Code or open a new session to register.
Using in Claude.ai, ChatGPT, Gemini, or other AI tools:
- Drag this file into a new chat, or copy and paste its contents
- Send your task or request as your next message — the AI will apply the skill where it fits
For Claude Code skill details: https://docs.anthropic.com/en/docs/claude-code/skills