Skip to content

skill.md

Latest commit

 

History

History
136 lines (92 loc) · 3.63 KB

skill.md

File metadata and controls

136 lines (92 loc) · 3.63 KB
Raw
name skill-name
description [One sentence: what artifact this skill produces]. Use when [trigger scenario 1], [trigger scenario 2], or [domain keywords].
metadata
apes-version category
1.1
product

[Skill Name]

Purpose

[One sentence: what engineering task this Skill completes and what artifact it produces.]

Input: [Required inputs from user or prior Skills]
Output: [Named artifact, e.g. "Validated Problem Statement document"]
Examples: See examples.md for worked input/output (optional).

Workflow

Step 1: Gather inputs

Collect the following from the user or context:

  • [Input 1]
  • [Input 2]
  • [Input 3]

If any required input is missing, stop and request it explicitly. Do not assume values.

Step 2: Analyze

[Describe analysis actions: map assumptions, identify gaps, compare alternatives.]

Step 3: Apply methodology

[Apply the named methodology: e.g. JTBD framing, RICE scoring, INVEST criteria.]

Step 4: Draft output

Produce the result using the Output Structure below.

Step 5: Validate

Run the Validation checklist. Fix all failures before presenting the final result.

Step 6: Recommend next steps

Suggest Skills from the Next Skills section based on the current outcome.

Decision Rules

Condition Action
Required input is missing Stop; list missing items; ask user
Input contains contradictions Surface contradictions; ask user to resolve before proceeding
Multiple valid approaches exist Present options with trade-offs; ask user to choose
Result fails Validation Revise and re-validate; do not deliver incomplete output
Scope exceeds single task Split into parts; complete only this Skill's scope

Validation

Before delivering the result, verify:

  • All required inputs were explicitly provided or confirmed
  • Output follows the Output Structure completely
  • No contradictions remain unresolved
  • No ambiguous terms without definition
  • Result is actionable without additional unstated assumptions
  • [Skill-specific criterion 1]
  • [Skill-specific criterion 2]

Anti-patterns

  • Assuming unstated context — inventing user goals, market data, or constraints not provided.
  • Role-play filler — generic advice not tied to the methodology.
  • Scope creep — implementing tasks that belong to Next Skills (e.g. writing a full PRD when only problem statement is requested).
  • Vague output — labels without measurable criteria ("improve UX", "increase engagement").
  • Skipping validation — delivering draft without running the checklist.

Best Practices

  • State assumptions explicitly in a dedicated section when user confirms them.
  • Use concrete examples from the user's domain, not generic placeholders.
  • Prefer tables and structured lists over prose paragraphs.
  • Reference the methodology by name and apply it consistently.
  • Flag risks and open questions in a separate section rather than hiding them.
  • Load examples.md when the user needs a reference output shape.

Output Structure

# [Artifact Title]

## Context
- **Source inputs:** [list]
- **Assumptions:** [list or "none"]

## [Main Section 1]
[Content]

## [Main Section 2]
[Content]

## Open Questions
- [ ] [Question requiring user input]

## Risks
| Risk | Impact | Mitigation |
|------|--------|------------|
| ... | ... | ... |

Next Skills

Outcome Recommended Skill
[Condition A] [category]/[skill-name]
[Condition B] [category]/[skill-name]
Default next step [category]/[skill-name]