- Home
- Skills
- Code Quality & Review
- ai-skills
ai Skills
A free open-source pack of skills that help coding agents work like experienced developers: evidence first, small blast radius, requirement anchoring, and checks against false or fake success. MIT licensed.
Free
Works with the AI Agents you already use
ai-skills
Example session with this skill installed
Add an export button on the orders page that downloads the current filtered order list as CSV. Keep the change small, pin a baseline first, look for evidence in the existing orders UI and API, and refuse a false-done if only a button stub exists without a working download.
- Read your context and instructions
- Compiled the ai-
- Generated the data export
Summary
Applied evidence-first + blast-radius + false-done before changing shared orders UI. Baseline pinned; export kept local to the orders page; refused to call the task done when only a stub button existed.
Baseline
- Repo commit:
a1b2c3d(orders page renders list + filters) - Existing evidence:
OrdersTablealready receives filtered rows; APIGET /api/orders?status=&q=returns the same filter set used by the UI - No existing CSV export helper in this route
Blast-radius plan
| Scope | Choice |
|---|---|
| Touch | app/orders/page.tsx, new exportOrdersCsv.ts helper next to the page |
| Avoid | shared Button primitive API, global download util, backend schema |
| Rollback | revert the two files; no migrations |
Evidence gathered
- Filter state lives in page URL query params — CSV must reuse those params, not a second client cache
- Order row shape includes
id, createdAt, total, status— enough for a useful CSV without new fields - No auth change required: same session cookie as the page
Result delivered
- Added
Export CSVbutton beside existing filters - Client builds CSV from the currently filtered rows already loaded for the table
- Empty state: button disabled with note when zero rows
- Manual check: filtered "paid" view downloads only paid rows
False-done refusal
An earlier draft only rendered a button with console.log('todo'). That is
not done: no file download, no filter reuse, no empty-state behavior. Kept iterating until a real CSV download worked on the filtered set.
Next steps for the human
- Run the project's existing orders UI test if present
- Spot-check one large filtered export in staging
- Do not treat "button visible" as success without a download
page.tsx
CSV · data export
Example file from a real run - the skill writes it into your workspace.
Connects securely to your tools. The creator never sees your data.
About this skill
What it does
A free open-source pack of process skills that help coding agents work like experienced developers: gather evidence before deciding, shrink change blast radius, lock requirements, and refuse false or fake success.
You can also install the same pack from the public GitHub repo: https://github.com/JeffChan702/ai-skills (MIT).
Skills included (SKILL.md)
- evidence-first: no decision without evidence
- blast-radius: shrink the blast radius before changing shared code
- requirement-anchor: lock requirements into a 12-cell matrix
- three-cards: turn a failure into three fill-in-the-blank questions
- false-done: an artifact existing is not the same as done
- context-guard: treat context as a workbench not a hard drive
- trustworthy-ai: make output trustworthy plus four check questions
- mail-html-lite: lightweight HTML for email bodies
Note: fake-success-guard is included as a script/README and may not load as a separate SKILL.md.
How it works
Load the SKILL.md skills into a coding agent that supports the SKILL.md standard. The agent then applies the process checks during planning, editing, and completion review.
Where it stops
These are process skills for coding agents, not a hosted service. They do not replace project-specific tests or human review.
Does not support greenfield development without baselines. Intentionally slows down workflow to ensure safety. Requires manual verification of AI-asserted evidence.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 2 days ago
- Free with an Agensi account
Permissions
Process SKILL.md skills guide the agent while reading and editing project files. Terminal is only needed if you run the optional fake-success-guard script from the pack.