Skills

August 2026 · Solo project · Instruction design and developer tooling

A set of labeled folders representing reusable agent skills

I started packaging my workflows so friends could use them without needing me to explain the same process each time. Some of them were running into problems I had already dealt with, so I turned my instructions, references, scripts, and design material into reusable skills.

A skill is closer to a specification than a prompt. It tells the agent when to use it, what process to follow, which material to load, and what a finished result should include. The same package can work across coding agents instead of living inside one conversation.

Why I packaged my workflows

A long prompt can describe what you want, but it often leaves the process open to interpretation. The agent may skip an important check, use a different method each time, or produce something that looks finished without completing the actual work.

My skills narrow that space. Each one handles a specific job and carries the instructions and supporting files needed to do it repeatedly. I built them from workflows I use myself, while also studying and adapting useful ideas from other public skills.

From my workflow to an installable skill

01Repeat the work

Use the same method often enough to know which parts matter.

02Write the contract

Define when the skill runs, what it does, and what finished means.

03Package the material

Add only the references, assets, scripts, and starters the job needs.

04Validate

Check the folder, frontmatter, name, description, and registry entry.

05Share

Publish through GitHub and install through the skills CLI.

The package begins with a workflow I already use, not an abstract prompt idea.

The six skills

The collection covers ADHD-friendly communication, presentations, editorial reports, launch films, technical diagrams, and reliable agent design. Friends use the ADHD, Presentation, and Editorial Report skills most often. I use the ADHD skill myself, especially in Codex with GPT models.

My I Have ADHD skill is adapted from Ayoub Ghriss's open-source project. It keeps responses short, leads with the action, replaces jargon with plain language, and breaks work into small steps. I removed the forced time estimates because I found them distracting. The other five packages turn my working methods into repeatable instructions with the files and checks each job needs.

The public GitHub repository for Mohtasham's reusable agent skills
The public repository keeps every skill, its supporting files, and validation workflow inspectable.
skillwhat it helps an agent do
i-have-adhd
Communicate clearlyShort answers, plain language, small actions, and visible progress
presentation
Build complete decksNarrative, slide design, output files, and rendered-slide checks
generate-editorial-report
Create evidence-led reportsFact gathering, editorial structure, figures, web, and PDF output
render-launch-film
Render launch videosA deterministic web timeline, MP4 renderer, stills, and contact sheets
draw-technical-diagrams
Draw precise SVG figuresComputed geometry, cutaways, callouts, hatching, and interaction
build-reliable-agents
Stop agent loops and stallsSmaller tool payloads, bounded retries, clear delegation, and probes
Each skill stays narrow enough to describe one job and the method for completing it.
Installationterminal
# Discover the collection
npx skills add mohtashammurshid/skills --list

# Check a local clone
python3 scripts/validate-skills.py

npx skills add mohtashammurshid/skills --skill presentation

Discover the packages, then install one skill or the full collection.

What makes a skill different

Every skill begins with a SKILL.md file that defines its trigger and workflow. More involved skills also include references, reusable assets, starter files, scripts, and checks. The agent loads the detailed material only when the job needs it.

The packages use the shared agent-skills format, so they can move between compatible coding agents. I do not have a growth target for the collection. I want it to remain useful. If one of my workflows saves someone from solving the same problem again, the package has done its job.

What lives inside a skill

Trigger
Name and descriptionTell the agent when this skill applies
Method
SKILL.mdWorkflow, constraints, and completion rules
Working material
references/Detailed guidance loaded only when needed
assets/Design systems, starters, and reusable files
scripts/Rendering, automation, and repeatable checks
Result
Finished artifactThe requested output, not only advice or an outline
ValidationEvidence that the package and its output still work
Simple skills may need only SKILL.md. Larger jobs carry the material required to produce and check the result.