Delegation
Problem Awareness, Platform Awareness, Task Delegation.
AI Fluency Skill Finder evidence
A personal project mapped to the 4D framework.
Skill Finder made my capability-search workflow more accountable: search existing options, inspect important files, score candidates, record risks, and ask for approval before installation.
Project evidence only. Certificate files are required before any certificate claim.
Description
Product Description, Process Description, Performance Description.
Discernment
Product Discernment, Process Discernment, Performance Discernment.
Diligence
Creation Diligence, Transparency Diligence, Deployment Diligence.
What problem it solved
When a task needs specialized help, an assistant can waste time by improvising a one-off helper while an existing skill, connector, documentation source, or workflow already exists. Skill Finder makes the search explicit and evidence-backed.
How it maps to the 4D framework
Delegation
Define the real task and decide which parts should be searched, inspected, scored, or left to human approval.
Description
Specify the output as an evidence-backed recommendation with source links, risks, confidence, and verification steps.
Discernment
Check accuracy, relevance, source freshness, dependency readiness, and fallback behavior before trusting the answer.
Diligence
Record what changed, keep installs under approval, and separate project-local rules from global agent rules.
What was validated
- Unit tests and static skill validation were recorded in the imported evidence set.
- The workflow changed after feedback when the first version was too narrow.
- The evidence files distinguish available tools from blocked or unavailable ones.
What still needs human approval
Installs, global mutation, auth or secrets, risky scripts, broad toolchain changes, and final trust decisions stay under human approval. The point is better judgment: know when a tool helps, know when it changes the environment, and leave the approval trail visible.
Boundaries
- Project evidence, not certificate proof.
- No official Anthropic project ownership claim.
- No production software claim.
- Focused on the evidence summary for this application version.