- Home
- Skills
- Testing & Debugging
- Proof of Fix
Proof of Fix
Turn a disputed bug fix into a reproducible before-and-after case, relevant regression results, and evidence tied to the files actually tested.
$9
Works with the AI Agents you already use
Proof of Fix
Example session with this skill installed
Fix parse_tags: empty or whitespace-only input should return an empty list; trim tag whitespace, ignore empty comma fields, and preserve order and duplicate tags. Use the same regression checks before and after the change. Show actual results.
- Read your context and instructions
- Compiled the proof of fix
Demonstration
Synthetic local parser fixture, actually executed during authoring. This is not a customer case or a cross-agent benchmark.
Verdict
Verified for the three stated parser requirements on this local fixture.
Reproduction
python3 -B -m unittest discover -s tests -v, from the demo project. Before the change, six tests ran and four failed: empty input, whitespace-only input, trimming, and empty comma fields. Order and duplicate preservation already passed. The original log and source were retained.
Cause and fix
parse_tags returned text.split(","), which retained blank fields and surrounding whitespace. The focused change strips fields and filters empty results, preserving their sequence and duplicates. The six regression tests were unchanged.
Final run
All six tests passed, exit code 0. Source and test fingerprints were collected before and after the run. The final manifest linked R1 (empty inputs), R2 (trimming and empty fields), and R3 (order and duplicates) to the passing run and its log. The helper returned consistent.
Drift probe
Adding a comment to parser.py after the run made the helper return inconsistent, identifying both source fingerprints as stale and leaving R1-R3 without current passing evidence. Restoring the exact tested bytes restored consistency. The drift result was retained separately.
Evidence and limits
Artifacts: evidence/before-fix.log, evidence/after-fix.log, evidence/final.json, evidence/consistent.json, evidence/drift-detected.json. Tested on macOS with Python 3.9.6. This validates the small parser case, not a whole application, other platforms, or all agent runtimes. Record consistency cannot prove that a test was meaningful or honestly reported. No commit or deployment occurred.
Connects securely to your tools. The creator never sees your data.
About this skill
A patch is easier to trust when you can repeat the failure, run the same case after the fix, and see exactly which files were tested.
Proof of Fix guides your coding agent through that process. It records acceptance criteria, separates application failures from broken test setup, preserves failed attempts, and runs relevant regression checks using your existing project tools. If the fix is already present and the original failure cannot be reproduced safely, the report says so.
An included, dependency-free Python helper fingerprints selected files and checks a final evidence manifest. It flags missing evidence, unpassed checks, uncovered requirement labels, and fingerprints that no longer match the current files. The helper reads explicit local paths; it does not run commands, contact services, install hooks, or modify your project. Your agent runs the actual tests using the host's tools.
You receive a scoped verification report: the case, change, actual checks and results, evidence paths, and what remains unverified. UI cases require real interaction outcomes; intermittent failures use bounded before/after samples.
Developed from verification and failure-handling patterns examined in G-Code. This is a portable workflow. The agent still judges whether the tests are meaningful. A consistent manifest is not independent proof of test execution, correctness or authenticity.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Testing & Debugging.
The helper checks declared records and explicitly watched files; it cannot authenticate logs, judge assertion strength or detect omitted dependencies. Missing prerequisites or checks leave verification incomplete. No runtime sandbox, automatic approval enforcement or correctness guarantee. Cross-agent and Windows/Linux integration have not been tested. No automatic commit, deployment or publication.
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 buying a skill.
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
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 4 days ago
Nothing to install. It runs in your AI agent.
- Optional
Python 3.9+ (evidence helper)
Use your existing project test tools. Python is optional for fingerprinting. Browser tools are needed only for UI checks. Existing project test and verification tools.
Permissions
Reads relevant project files and writes buyer-owned evidence or tests. Application code is edited only when repair is requested. The helper itself is read-only and uses no network. No connected service is required by the skill; project checks may have their own dependencies.