All skills
jwynia avatar

/competency

@6ac6c30
by J Wyniajwynia/agent-skills160 stars
20

Guide competency framework development and operation. Use when building training that produces capability, when existing training doesn't produce competence, when structuring knowledge for multiple audiences, or when setting up feedback loops to surface gaps.

  • 5 files
  • 30.9 KB
  • MIT
  • Updated 8 months ago
  • GitHub

Use this Skill: https://skilld.dev/gh/jwynia/agent-skills/competency

This session only. Nothing lands on disk.

templatesscenario.md

≈879 tokens on demand. Your agent reads this file only when SKILL.md points to it.

Scenario Template

Use this template to design scenarios that test competencies through realistic situations requiring judgment.

Template

### Scenario: [Descriptive Name]

**Core decision structure:** [What judgment or trade-off is being tested]

**Interview variant:**
> [Generic situation that doesn't require knowledge of specific organization. Tests reasoning approach.]

**Assessment variant:**
> [Organization-specific version using real tools, policies, team names. Tests application of learned content.]

**Ongoing variant:** (optional)
> [Real situation that occurred, anonymized if needed. Tests continued competence.]

**Competencies assessed:** [Competency IDs this scenario tests]

**What good looks like:**
- [Key consideration a competent response would include]
- [Another consideration]
- [What distinguishes competent from strong]

**Red flags:**
- [What a weak response would miss]
- [Common mistakes]
- [Misconceptions revealed]

Scenario Design Checklist

  • Realistic: Situation people will actually encounter
  • Incomplete information: Like real life, not everything is specified
  • Requires judgment: Can't be answered by searching documentation
  • Gradable: Has better and worse responses, not just right/wrong
  • Time-efficient: Can be evaluated in reasonable time
  • Variant-ready: Interview version doesn't require org-specific knowledge

Testing Your Scenario

  1. Too easy: If everyone gets the same answer quickly, add ambiguity
  2. Too hard: If no one can engage meaningfully, provide more context
  3. Not testing competency: If responses don't differentiate by competency level, redesign
  4. Lookup-able: If someone could ctrl+F the answer in docs, it's testing recall not judgment

Example: Tool Evaluation Scenario

### Scenario: Third-Party Tool Evaluation

**Core decision structure:** Given incomplete vendor information, identify what you need to know and how to verify claims.

**Interview variant:**
> A team wants to use a third-party AI coding assistant. They've sent you the product page. What questions do you need answered before you can evaluate it? How would you find those answers?

**Assessment variant:**
> Team X wants to use [SpecificTool]. Here's their use case: [description]. Here's the vendor's security documentation: [link]. Evaluate this against our data policies and write up your recommendation.

**Competencies assessed:** DP-1 (Data classification), DP-2 (Service model differentiation), DP-4 (Vendor evaluation)

**What good looks like:**
- Asks about data retention, training data usage, processing location
- Distinguishes marketing claims from verifiable commitments
- Identifies what data the team would actually send
- Recognizes gaps in available information
- Knows when/how to escalate uncertainty

**Red flags:**
- Takes marketing claims at face value
- Doesn't ask about data flow
- Binary yes/no without nuance
- Can't articulate what would change the assessment

Scoring Scenarios

Level Meaning Indicators
Not demonstrated Didn't engage meaningfully Missed obvious considerations, couldn't articulate approach
Partial Some considerations, gaps in reasoning Got some factors, missed important ones
Competent Addressed key considerations Sound reasoning, could do independently
Strong Sophisticated judgment Identified non-obvious factors, could teach others

Source: SKILL.md on GitHub

No third-party reports yet.

Signed by skilld at 6ac6c30. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub 2 months ago.

Dormantupdated 8 months ago
Other metadata
metadata
{
  "author": "jwynia",
  "version": "1.0",
  "domain": "competency",
  "cluster": "methodology",
  "type": "diagnostic",
  "mode": "assistive"
}

README badge

README badge for jwynia/agent-skills/competency