half-fixed: find bug fixes that missed sibling code

    1

    Finds bug fixes that reached one adapter, driver or provider but not its siblings, then checks each one before changing code.

    Free

    1 useSecurity scanned

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    half-fixed: find bug fixes that missed sibling code

    Example session with this skill installed

    Find half-fixed bugs in this repo, focusing on the tool maps.

    • Read your context and instructions
    • Compiled the half-fixed

    I ran the bundled scanner on prism-php/prism with scan --group ToolMap --json and got one high-confidence finding.

    src/Providers/*/Maps/ToolMap.php
      hf-5505660  high  fixed in 2 of 10, still old in 4
        fixed    DeepSeek, OpenRouter 9447c71 2025-10-14 fix(DeepSeek|OpenRouter): Fix exceptions during tool conversion without parameters (#661)
        missing  Groq:21, Mistral:21, XAI:21, Z:21
        + ...$tool->hasParameters() ? [
        + ] : [],
    

    Triage

    IDGroupMissingVerdictReason
    hf-5505660Maps/ToolMap.phpGroq, Mistral, XAI, Zgap9447c71 stopped DeepSeek and OpenRouter failing on tools with no parameters. Groq, Mistral, xAI and Z still send "properties": [] for those tools, which their APIs reject.

    What I checked: the original commit and its PR (#661), the code at line 21 in each of the four missing tool maps, and whether any of them guard against empty parameters somewhere else. None do. The fix only wraps the existing block, so the same change applies to each sibling.

    I haven't changed any code. If you want, I'll fix the four tool maps with a failing test first, then run the project's own checks, and draft a PR that credits 9447c71. (This finding was fixed upstream in prism-php/prism#1038.)

    Connects securely to your tools. The creator never sees your data.

    About this skill

    When a bug is fixed in one provider adapter, database driver or cache store, the siblings often have the same code and the same bug. half-fixed finds those gaps in your git history.

    What it does

    • Runs a bundled scanner (scripts/half-fixed.js) over git history. It groups sibling files, replays each commit that changed only some of them, and reports siblings that still have the old code.
    • Has your agent triage every finding before touching code: gap, intentional, handled-elsewhere, obsolete or unsure, each with evidence from the original fix.
    • Reports a short table. It only fixes things if you ask, one finding at a time, with a failing test first.

    Why trust it

    • Zero dependencies. Needs Node 18+ and git, nothing else. Runs offline.
    • Never executes the scanned repo's code and never writes to the repo during a scan.
    • Treats everything in the repo (commit messages, code, config) as data, not instructions.
    • Fast: 5,000 LiteLLM commits in 69 seconds.
    • Honest about accuracy: on LiteLLM, 13 of 66 high-confidence findings were real gaps. The rest were deliberate or already handled, which is why the agent checks each one.
    • Has found real bugs, fixed upstream in prism-php/prism#1038, BerriAI/litellm#44925 and BerriAI/litellm#44936.

    Try it Ask your agent: "Find half-fixed bugs in this repo." Scanner only: npx half-fixed scan

    MIT licensed. Source, evidence and full docs: https://github.com/heyzainali/half-fixed

    Requires Node 18+ and a full (non-shallow) git history. Cannot distinguish between missed fixes and intentional architectural differences without manual triage.

    How to use

    1. 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. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 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

    1 use

    Times used to date

    Free forever

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 29 days ago

    • Free with an Agensi account

    Requires

    Node.js 18+
    git
    Node.js
    on

    Permissions

    Terminal / Shell
    Read Files
    Write Files
    Listed29 days ago

    What's inside

    Frequently Asked Questions

    The skill identifies 'half-fixed' bugs where a developer applied a fix to one file (like a MySQL driver) but forgot to apply the same change to its siblings (like the PostgreSQL or SQLite drivers). It prevents technical debt by ensuring consistency across parallel implementations.

    This skill is compatible with any AI agent that can execute Node.js scripts and access a local terminal. It works best with agents capable of reading git logs and file contents to interpret the scanner's JSON output.

    No, the scanner is language-agnostic. It detects patterns, shared suffixes, and structural similarities in your file paths, making it effective for everything from hardware abstraction layers in C++ to cloud adapters in TypeScript.

    You receive the core Node.js scanner script, the specialized agent instructions for interpreting git history, and documentation on how to integrate the scan results into your automated code review workflow.

    The scanner requires a non-shallow git clone to analyze history effectively. If your CI/CD environment or agent uses a shallow clone (depth=1), it may not have enough historical context to identify when a sibling was bypassed.

    New in Code Quality & Review