Works with the AI tools you already use
Agent Capability Router — Tool, Script, MCP, or Skill?
"Should this be a skill or an MCP?" is the wrong question. Here's the right one.
$12
Agent Capability Router — Tool, Script, MCP, or Skill?
Example session with this skill installed
My agent needs to generate weekly PDF reports from our Jira board. It has Jira access but the formatting is always messy. Should I make this a tool or a skill?
- Read your context and instructions
- Compiled the agent capability router
- Generated the document
The agent can reach Jira but does the job wrong. This is a handbook problem, not a badge problem.
Routing Decision: Skill with Bundled Script
You should implement this as a Skill containing a deterministic Python/Bash script for the PDF generation.
| Component | Recommendation | Context Cost |
|---|---|---|
| Mechanism | Skill (SKILL.md + report_gen.py) | ~100 tokens (startup) |
| Logic | Bundled Script | 0 tokens (logic stays in file) |
| Access | Existing Jira Auth | N/A |
Reasoning:
A Tool definition sits in your context every single turn, costing you "rent" for a task you only do once a week. By using a Skill, the model only loads the formatting instructions when you ask for the report. By pushing the "messy formatting" logic into a bundled script (e.g., using a library like fpdf or reportlab), you ensure the output is identical and professional every time without the agent having to "guess" the layout at runtime.
Next steps
- Create a new skill folder named
jira-weekly-reports. - Move the PDF formatting logic into a standalone script within that folder.
- Write a
SKILL.mddescription that triggers specifically on Jira reporting keywords.
agent-capability-router-tool-script-mcp-.pdf
PDF · document
Example file from a real run - the skill writes it into your workspace.
Connects securely to your tools. The creator never sees your data.
What you get
About this skill
Every capability you add to an agent is one of five mechanisms: a tool, a bash script, code written at runtime, an MCP server, or a skill. Pick wrong and you either pay context rent on every single turn for something used twice a month, or you bolt on more access to fix a problem access was never going to fix.
The badge / handbook test
A new hire needs a badge to get through the door and a handbook to learn how the job is done here, and they are not interchangeable.
- If the agent couldn't REACH the thing, that's a badge problem and more instructions won't help.
- If it reached the thing and did the job WRONG, that's a handbook problem and more access won't help.
What you get
It prices each access mechanism by WHEN it costs context, not by what it can do. Returns exactly one mechanism, the context cost you're accepting, and whether heavy logic should be pushed into a bundled script (where the code never enters context at all — only its output does). One routed answer, never a comparison menu.
Use cases
- Your agent keeps getting the same task wrong — decide in one pass whether that's an access gap or a missing handbook
- Deciding whether an integration justifies building or connecting an MCP server
- Auditing an agent stack that's gotten slower with every capability added, to find what's paying context rent
- Explaining to a non-technical client why the connection is the cheap part and writing down how their shop works is the part worth paying for
How to install
Works the same in every agent - Claude, Cursor, Codex, Copilot and 20+ more.
- 1
Download the ZIP
Free skills download straight away. Paid skills unlock right after purchase.
- 2
Unzip into your skills folder
Every agent reads skills from one folder on your machine. Drop the unzipped folder in there.
- 3
Ask your agent to use it
Restart the agent if it was already running. It picks the skill up automatically - no config needed.
Skills folder by agent
Click the path to copy it. Create the folder if it does not exist yet.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill passes our trust checks below.
Security scanned
Passed our 8-point scan before listing
Fresh listing
Recently published to Agensi
30-day refund
Not a fit? Get your money back
Trust & safety
Security scanned
Verified clean 6 days ago
- Passed all security checks, Safe to install