- Home
- Skills
- Testing & Debugging
- Finish the Wiring
Works with the AI Agents you already use
Finish the Wiring
Fix features whose parts exist but don't connect. Trace the real path, repair the break, and verify what actually works.
Free
Finish the Wiring
Example session with this skill installed
Use Finish the Wiring on this local Python profile CLI. The profile CLI says Saved after I run save --name Alice, but a fresh show command still displays the old name. Find and fix the broken connection. Empty names should fail without claiming success or changing stored data.
- Read your context and instructions
- Compiled the finish the wiring
Finish the Wiring: local profile CLI demonstration
This is a synthetic local fixture used for a marketplace demonstration. It is not real customer proof or a general benchmark. No network calls were used.
Result
Reproduced the reported bug: save --name Alice printed Saved, but a new show process still read Sam.
The broken connection was in profile_cli.py's save branch: it changed the loaded dictionary without calling the existing storage.save_profile. The repaired branch calls that function at line 18 and prints its returned persisted profile at line 21. Validation and write errors return a nonzero exit before success is printed; a missing name also returns a clear error. The existing storage implementation was unchanged.
Verified with actual CLI subprocesses
- Saving Alice is visible to a fresh
showprocess. - The response matches the normalized persisted value.
- Empty and whitespace-only names fail with no
Savedoutput. - Failed saves leave the stored file byte-for-byte unchanged.
- Missing
--namefails without modifying storage. - The unrelated
theme: darksetting survives successful and failed saves.
verify_profile_cli.py is the focused, repeatable regression check. It copies the two implementation files into a temporary directory beneath this demo folder, uses real CLI processes, asserts the outcomes, and removes its temporary fixture. The main demonstration's profile.json now contains Alice with the original dark theme.
All requested checks passed. No blocker remains for this fixture. This verifies only the profile save path; concurrency, crash-safe writes, and unrelated application behavior were not tested.
Actual terminal evidence: before the repair
Initial profile
> py -X utf8 profile_cli.py show
exit: 0
{"display_name": "Sam", "theme": "dark"}
Reported save
> py -X utf8 profile_cli.py save --name Alice
exit: 0
{"status": "Saved", "profile": {"display_name": "Alice", "theme": "dark"}}
Fresh process after save
> py -X utf8 profile_cli.py show
exit: 0
{"display_name": "Sam", "theme": "dark"}
Actual terminal evidence: after the repair
Save through repaired CLI
> py -X utf8 profile_cli.py save --name Alice
exit: 0
{"status": "Saved", "profile": {"display_name": "Alice", "theme": "dark"}}
Read in a fresh process
> py -X utf8 profile_cli.py show
exit: 0
{"display_name": "Alice", "theme": "dark"}
Focused process-level regression checks
> py -X utf8 verify_profile_cli.py
exit: 0
$ py -X utf8 profile_cli.py save --name " Alice "
exit: 0
stdout: {"status": "Saved", "profile": {"display_name": "Alice", "theme": "dark"}}
stderr: <empty>
$ py -X utf8 profile_cli.py show
exit: 0
stdout: {"display_name": "Alice", "theme": "dark"}
stderr: <empty>
PASS: separate process reads Alice; normalization and dark theme preserved.
$ py -X utf8 profile_cli.py save --name ""
exit: 1
stdout: <empty>
stderr: Error: Display name must not be empty
$ py -X utf8 profile_cli.py show
exit: 0
stdout: {"display_name": "Alice", "theme": "dark"}
stderr: <empty>
PASS: empty name fails without success output; stored bytes unchanged.
$ py -X utf8 profile_cli.py save --name " "
exit: 1
stdout: <empty>
stderr: Error: Display name must not be empty
$ py -X utf8 profile_cli.py show
exit: 0
stdout: {"display_name": "Alice", "theme": "dark"}
stderr: <empty>
PASS: whitespace name fails without success output; stored bytes unchanged.
$ py -X utf8 profile_cli.py save
exit: 2
stdout: <empty>
stderr: usage: profile_cli.py [-h] [--name NAME] {show,save}
profile_cli.py: error: --name is required for save
$ py -X utf8 profile_cli.py show
exit: 0
stdout: {"display_name": "Alice", "theme": "dark"}
stderr: <empty>
PASS: missing name fails without success output; stored bytes unchanged.
All profile CLI regression checks passed.
Connects securely to your tools. The creator never sees your data.
About this skill
Make the button actually do what it says
A Save button reports success, but the value disappears after a reload. A form looks complete, but never reaches its handler. An API returns the right data, but the screen stays stale.
Finish the Wiring guides your coding agent through one feature's full path: the user action, handler, service or API, storage when relevant, and displayed result. It looks for the broken connection, makes a focused repair using your project's existing patterns, and checks the resulting behavior.
Start here
Already connected to Agensi? Paste this one line into your coding agent with your project open:
Use the free Agensi skill "Finish the Wiring" by bhunesh bansal to trace, repair, and verify [feature] in this project.
Replace [feature] with what is broken, for example: "the profile Save flow; the name resets after refresh".
First time? Connect Agensi once
Choose the command for your agent and run it in your terminal. These commands connect Agensi; the prompt above loads this skill.
Claude Code
claude mcp add --transport http agensi https://mcp.agensi.io/mcp/auth
Then open Claude Code, run /mcp, and choose Authenticate for Agensi.
Codex CLI
codex mcp add agensi --url https://mcp.agensi.io/mcp/auth
Then run codex mcp login agensi.
Complete the browser sign-in to Agensi and click Allow. Then paste the use prompt above. If Agensi is already connected, skip setup. No separate skill package installation is needed through Agensi.
What you get
- A trace of the requested feature and the specific broken connection.
- A focused code repair that preserves unrelated work and intentional demos.
- Verification of the actual result when the environment is available, including a fresh read or reload for persistence claims.
- A short report of the changes, checks completed, and remaining blockers.
Try it
"Use Finish the Wiring on the profile Save feature. Changes disappear after refresh. Find the broken connection, fix it, and verify the result."
Add "Review only; do not edit files" if you only want findings.
Scope
One small instruction-only SKILL.md for coding agents such as Claude Code and Codex. It follows your existing project contracts and does not require a specific framework. Repairs and runtime checks depend on the agent's tools and access to your project. It does not supply a backend, invent missing requirements, or authorize deployment or production writes.
Covers one requested feature, not the entire application. Missing services, schemas, credentials, or product decisions can prevent completion. Runtime verification requires the relevant project environment and tools; otherwise the agent must distinguish static inspection from observed behavior. Existing project dependencies and service costs still apply. No guaranteed fixes, deployment, or production writes.
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 23 days ago
- Free with an Agensi account
Requires
Permissions
Read the relevant project files and edit only the requested feature when repair is requested. Run project commands/tests in a suitable local or test environment. Browser/network access may be used for the selected local/test flow; no mandatory external account or MCP connection. Review-only requests do not authorize edits. No deployment or production writes.