feedback

# Skill: Feedback ## When Not to Use - When the task is to rewrite or produce new content rather than evaluate existing content — produce the new version first, then apply this skill - When the work is a final legal, compliance, or regulated deliverable requiring professional review — structured feedback is appropriate, but this skill cannot substitute for specialist sign-off - When unsolicited critique would be inappropriate given the context (e.g., personal creative work shared for enjoyment, not evaluation) --- ## Interaction Protocol **Before starting**, ask if not already clear: 1. What is the purpose of the work — what is it trying to achieve, and for what audience? 2. Are there specific concerns to prioritise, or should the review be comprehensive? 3. What level of critique is appropriate — high-level structural, line-level, or both? **Output style**: - Group findings by category (Structure, Argument/Logic, Clarity, Accuracy, Completeness, Style) - Assign each finding a priority: `Critical`, `Major`, `Minor`, or `Suggestion` - For every finding: state the specific problem, identify its location or instance, explain the consequence, and give a concrete recommended change - End with a summary: total findings by priority, the work's key strengths, and the single highest-leverage improvement - Do not produce generic encouragement — observations that do not identify a specific problem or produce an actionable recommendation are not findings --- ## Inputs and Outputs **Input**: The work to be reviewed (text, argument, plan, decision, or document), with optional context: stated purpose, audience, and specific concerns **Output**: Structured findings grouped by category, each with priority, location, problem, consequence, and recommendation; overall summary with key strengths **Composability**: Use after strategy-author, research, or technical-writer (to evaluate their output); use alongside citation-discipline (to assess sourcing quality); use before remove-ai-slop (to identify structural issues before prose cleaning) --- ## Purpose Produce specific, evidence-grounded feedback that enables the author to improve their work in concrete ways. Every finding must be tied to an observable problem in the work — not preference, taste, or convention for its own sake. --- ## 1. Review Categories Apply all relevant categories. Skip a category only if it is entirely inapplicable to the work type. ### 1.1 Structure - Does the work have a clear beginning, middle, and end appropriate to its type and purpose? - Is the order of information logical — does earlier content provide what later content requires? - Are sections proportionate to their importance? - Is anything present that serves no structural function? ### 1.2 Argument and Logic - Is the central claim or thesis clear and explicitly stated? - Do the supporting points actually support the central claim? - Are there gaps in the reasoning where steps are skipped or assumed? - Are there contradictions between different parts of the work? - Are conclusions supported by the evidence presented, or do they exceed it? ### 1.3 Clarity - Would a reader in the stated audience understand this without additional context? - Are key terms defined where necessary? - Are ambiguous pronouns, vague references, or unclear antecedents present? - Are sentences or paragraphs doing too much — trying to communicate multiple distinct ideas at once? ### 1.4 Accuracy - Are factual claims correct and appropriately supported? - Are there statements that overstate or misrepresent the evidence? - Are assumptions stated explicitly, or are they embedded without acknowledgment? ### 1.5 Completeness - Does the work address what it sets out to address? - Are there obvious objections, alternatives, or failure modes not acknowledged? - Are necessary prerequisites, context, or definitions missing? ### 1.6 Style and Tone - Is the tone appropriate for the audience and purpose? - Is the language register consistent throughout? - Are there patterns that reduce impact: hedging, filler phrases, passive constructions where active would be stronger? --- ## 2. Priority Classification Assign each finding one of the following: | Priority | Criteria | |---|---| | `Critical` | The work fails its purpose or misleads the reader if this is not addressed | | `Major` | The work is substantially weakened, confusing, or incomplete without this fix | | `Minor` | The work would be meaningfully improved, but functions without the change | | `Suggestion` | An optional improvement that may or may not fit the author's intent | --- ## 3. Finding Format Each finding must contain: ``` **[PRIORITY] Category — Short title** Location: <section, paragraph, sentence, or specific quote> Problem: What is wrong and why it is wrong. Consequence: What fails or degrades for the reader if this is not addressed. Recommendation: Specific change to make. ``` --- ## 4. Summary Format End every review with: ``` ## Summary Findings: X Critical, X Major, X Minor, X Suggestions **Strengths**: [Two or three specific things the work does well — must be observable, not generic praise] **Highest-leverage improvement**: [The single change that would most improve the work's effectiveness] ``` --- ## 5. Consistency Check Before finalising, verify: - Every finding has a specific location — no vague "the argument is weak" statements - Every recommendation is concrete — no "consider improving" without specifying how - Priority levels are consistent — the same class of problem is not rated `Critical` in one section and `Minor` in another - Strengths identified are genuine and specific, not formulaic reassurance - Summary counts match the body --- ## 6. Failure Modes - Generating generic encouragement that dilutes the signal of genuine findings - Flagging stylistic preferences as structural failures - Assigning `Critical` priority without demonstrating a concrete failure for the reader - Producing findings without a recommended fix - Omitting the `Strengths` section — honest feedback includes what is working - Letting the summary contradict the findings (e.g., calling work "strong" when Critical findings exist)