Skip to content
Muhammet Şafak
tr
Cover — a terminal reading 'Fix proposed as sentryfixer_output.diff' and 'Regression test created in tests/test_fix.py', the diff open beside it, a violet SentryFixer keycap

SentryFixer: reading a Sentry error against the code and drafting the fix

How a Claude Code skill reads a Sentry issue against the working tree and drafts the fix as a diff and a regression test, touching no file before approval.

What it does today

Reads a Sentry issue against the working tree, produces a root-cause analysis and a fix plan, and touches no file without explicit approval. Anyone using Claude Code who needs to triage a production error can install the skill.

Started
July 2026 — September 2026
Left Labs

Form

Open-Source

Technologies

Python Bash Sentry API sentry-cli MCP Claude Code

The gap between error monitoring and the fix always sits in the same place: Sentry tells you what blew up, but working out how that line looks today, whether the code path changed since the error, and what a fix might break is still on you. SentryFixer is the Claude Code skill I wrote to automate that middle step.

Four steps from event to fix plan

Given a Sentry issue, the skill runs through:

  1. Compressing 50–300 KB of event data into a redacted summary.
  2. Mapping stack trace frames onto the real source files in the working tree.
  3. Checking freshness — whether the failing code changed since the error fired.
  4. Reasoning about the root cause and proposing a concrete fix with tests.

The strict part comes last: no file changes until the plan is approved. No automatic commits, no git operations beyond read-only history. Once approved, the diff is applied, the tests run, and the result is reported as-is.

Narrow dependencies, hard approval gate

The dependency surface is deliberately narrow: Python 3.8+ with the standard library only, plus curl/bash. There are three ways into Sentry — the API directly, sentry-cli or the Sentry MCP server — whichever is available. MIT licensed, repository public.

Two decisions sit at the centre of the design. First, context compression: raw event data never goes to the model as-is; it is reduced to the frames that matter, and personal data is redacted at that step. Second, the approval gate: the skill’s write access stays shut until someone reads the plan and says go.

The limit of the freshness check

The open question on the bench: freshness currently checks whether the file changed. What I actually want is a verdict on how much the diff between the error’s commit and today’s code touches the root cause. That is the next round.

Tags: #ai#debugging#automation
Share:

More Records

All records

A CLI that solves a problem repeatedly in isolated sessions and measures how diverse the solution paths are, with JSON and HTML reports. Not on PyPI.

What it does today

Has a model solve the problems of a suite YAML repeatedly in isolated sessions, measures how diverse the solution paths are and how much accuracy drops when a problem is changed on the surface, writes a JSON and HTML report per problem and fails CI on suspected memorization. Engineers choosing a model or checkpoint, or producing fine-tuning data, can clone the repository and install it with uv; it is not on PyPI.

Open-Source Python Typer Pydantic +3 more
November 2026 — Ongoing

A Python library I wrote to get schema-valid output from small on-device language models; its first reference benchmark missed some targets.

What it does today

Takes a Pydantic model or a JSON Schema and gets schema-valid output from a local model: grammar (llama.cpp only), deterministic repair, re-ask and a cloud fallback that is off by default (Claude, OpenAI, Gemini) work; there is a metrics collector, cost estimation and an OTLP observer. The MVP is complete but no release is published; in the first reference benchmark two models missed the targets on the ONNX path. Python developers embedding a local model in an application can try it by cloning the repository; it is not on PyPI.

Open-Source Python Pydantic llama.cpp +2 more
November 2026 — Ongoing

A proxy between an agent and its MCP servers: scans results for injection, masks PII, enforces access and logs decisions to a hash-chained audit log.

What it does today

Sits between an agent and its MCP servers over stdio (`wrap`) or HTTP (`serve`), scans tool results and resource contents for prompt injection, masks personal data, enforces access by tool, role and dangerous capability combination, pins a server's tool definitions and writes every decision to a hash-chained audit log. Anyone wiring MCP servers into agents can try it by cloning and building the repository; it is not published on npm.

Open-Source TypeScript Node.js MCP +2 more
September 2026 — Ongoing

Search the site

Start typing to search posts, projects and pages.

Esc to close Powered by Pagefind