output-format

Pure-markdown output rules for plans, reports, reviews, and any structured agent output.

SKILL.md

GitHub

Output Format

Use this format for ALL structured output — plans, reports, reviews, diagnostics, conformance reports, post-implementation reviews, and any other agent-produced documents. It renders cleanly on any platform (Claude Code, opencode, Codex CLI, ChatGPT, any terminal or web UI).

Readability Rule (Main Window)

Everything this skill governs is read directly by a human in the main chat window — not a file on disk. Write for light, fast reading, not dense prose:

  • Short paragraphs — 2–4 lines max. Break up anything longer.
  • Lead with the point — conclusion or answer first, supporting detail after.
  • Bullets and tables over prose whenever there are 3+ related items, options, or steps.
  • No padding — cut restated context and filler transitions ("It's worth noting that…").
  • Headers only when the content needs navigation — a two-paragraph answer doesn't need section headers.
  • One idea per sentence.

This also cuts output tokens: a padded answer costs more to generate than a tight one saying the same thing. Does not apply to files written to disk (code, docs, ADRs) — those follow their own format's conventions.

Core Rules

  • Pure markdown only — no HTML, no raw ANSI, no escape sequences
  • No box-drawing Unicode — never use ┌ ─ ┐ │ └ ┘ ┃ ━ ┏ ┓ ┗ ┛ ┣ ┫ ┳ ┻ ╋ ═ ║ ╔ ╗ ╚ ╝ ╠ ╣ ╦ ╩ ╬
  • No decorative Unicode symbols — never use ☐ ☑ ☒ ✓ ✗ ★ ☆ ✦ ✧ ❗ ❓ ➜ → ⇒ ⇨ · • ◦ ‣
  • No emojis for decoration — use ✅ ❌ ⚠️ 📌 🔄 only when they carry real semantic meaning (completion, error, warning, action item, pending). Never use emojis purely decoratively.
  • Markdown tables — use |---|---| syntax, never ASCII box-drawing tables

Section Structure

## Title

### Subtitle

Body text with **bold** for emphasis and `inline code` for paths, commands, or filenames.

- List items with `-` for unordered lists
- Use `1.` for ordered lists

Code blocks with language tag:

```language
code or structured data

## Report Template

Use this structure for reports, reviews, and diagnostics:

[Report Type] — [Brief Title]

Summary

[1-3 sentences: overall verdict and key takeaway]

[Section Name]

  • [LABEL] [item] — [description and action]
    • Location: path/to/file.ext:line
    • Recommendation: [what to do]

[Next Section]

  • ...


### Severity Labels

Use these consistently:

| Label | When |
|---|---|
| `[BLOCKING]` | Must fix before proceeding |
| `[HIGH]` | Significant issue, should fix soon |
| `[MEDIUM]` | Moderate concern |
| `[LOW]` | Minor, take it or leave it |
| `[INFO]` | Information only, no action required |
| `[SUGGESTION]` | Improvement, not a blocker |
| `[QUESTION]` | Needs clarification |

## Conformance Report

Conformance Report — [Sprint / Date]

Summary

[1-2 sentences]

Deviations

File / Area Deviation Severity Action
path/to/file.ext [what] [severity] [fix]

Tech Debt Tracked

Item Introduced Owner Target

Conformant Areas

  • [area that passed review]

## Post-Implementation Review

Post-implementation review

Code review (critical only)

  • [bullets]

QA (gaps / risks)

  • [bullets]

Summary

[1-2 sentences: verdict and recommended next step]


## Security Review

Security Review — [Brief Title]

Executive Summary

[2-3 sentences: overall risk posture]

Critical Findings

  • [CRITICAL] [title]
    • Location: path/to/file.ext:line
    • Description: [what the vulnerability is]
    • Attack scenario: [how an attacker would exploit this]
    • Remediation: [specific fix]

High Findings

  • [HIGH] [title]
    • ...

Medium / Low / Info

  • [MEDIUM / LOW / INFO] ...

## Plan Template

Plan — [Task Name]

Context

[One or two sentences: what triggered this task and why it matters]

Scope

In scope:

  • [what will be changed]

Out of scope:

  • [what will NOT be touched]

Approach

[One paragraph: chosen strategy and reasoning]

Steps

# Action Files Complexity Par.
1 [action] path/to/file [Low/Med/High] [A/B/—]

Complexity: Low=routine, Medium=multiple touch points, High=architectural Par.: same letter = parallel, — = sequential

Risks & Dependencies

  • Risk: [description] — Mitigation: [how handled]
  • Depends on: [prerequisite]
  • Assumption: [assumed state]

Definition of Done

  • [acceptance criterion 1]
  • [acceptance criterion 2]
  • Linter / type-checker passes
  • Tests pass (if applicable)
  • Documentation updated (if applicable)

Awaiting your approval before proceeding. Reply "approved" to execute or provide feedback to adjust.


## Diagnostics

[Diagnostic Type] — [Brief Title]

Status

[PASS / FAIL / WARNING] — [one-line summary]

Details

  • Check: [what was checked]
    • Expected: [value]
    • Actual: [value]
    • Result: [PASS / FAIL]

Recommendations

  • [if applicable]