RepoWitness

The description states that RepoWitness audits Markdown postmortems against a local Git repository. For each remediation claim, it returns one of three outcomes:

• supported by repository evidence;
• no supporting evidence found within the completed search scope;
• not testable from the available evidence.

It records paths searched and excluded, Git range, tools executed, failures, evidence anchors, and provenance. A supported result must include a mechanically verified file, line, and commit anchor.

The tool is described as never turning “I did not find it” into “it does not exist,” nor inferring deployment or runtime state from Git.

Evidence The author’s own write-up.

Inference This is a read-only audit tool for verifying Git-based changes in incident reports, with deterministic logic to avoid false positives.

The author states that RepoWitness was built because “a postmortem action item can be marked ‘Done’ without anyone checking whether the repository contains evidence of the promised change.” The tool aims to make this gap inspectable instead of treating workflow status as truth.

It is positioned as a witness to what Git repositories can and cannot support, based on its own self-description.

Evidence The author’s own write-up.

Inference The product is framed as a solution to a specific problem in incident management workflows — the disconnect between ticket state and actual code changes.

The description does not name specific customers or personas. However, it implies that RepoWitness targets developers or teams working with Git repositories and postmortems, particularly those involved in incident response or reliability engineering.

Evidence The author’s own write-up.

Inference Likely users are internal engineering teams or SREs who manage incidents and want to verify that remediation steps were actually implemented.

There is no evidence of a business model, pricing structure, or monetization strategy in the description. The tool is presented as a static demo with no API keys required, nor any indication of paid services or subscriptions.

Evidence The author’s own write-up.

Inference No commercial model is evident from the self-reporting content.

The tool was built using Codex (a language model) for design and implementation assistance. It uses deterministic host code to verify evidence anchors, enforce search scope boundaries, and retain final verdict authority. GPT-5.6 Sol was used once for narrow semantic assessment over fixed candidates in a synthetic fixture — it did not access the repository or issue final verdicts.

The fastest judge path is an API-key-free static demo that can be opened directly from demo/index.html.

It includes features like:

• Bounded read-only file and Git tools;
• Evidence packet generation;
• Anchor verification;
• Search-scope accounting;
• Provenance records;
• Deterministic verdict gate;
• Replay system;
• Security regressions;
• Browser QA;
• Release checks;
• Static evidence board.

Evidence The author’s own write-up.

Inference The tool is designed for deterministic, secure evaluation of Git-based claims with minimal external dependencies.

The description mentions:

• Synthetic demo results where deterministic evaluation matched all 5 expected outcomes.
• One approved live GPT-5.6 semantic micro-smoke with 0 false SUPPORT results and 0 fabricated anchors.
• Final main branch passed 206 tests, including desktop and mobile browser QA.

However, the author explicitly notes that these are limited synthetic results, not a real-world benchmark, and do not establish general accuracy or production-state verification.

There is no evidence of customers, revenue, usage metrics, or adoption beyond the author’s own testing.

Evidence The author’s own write-up.

Inference The tool exists in prototype form with some internal validation but lacks external traction or real-world deployment data.

The description does not mention any competitors. It is unclear whether similar tools exist in the market for Git-based incident verification or postmortem auditing.

Evidence The author’s own write-up.

Inference No competitive landscape is described, so it's unknown if this addresses an existing gap or introduces a new category.

• No real-world usage or adoption: All evidence points to internal testing and synthetic validation.
• Unproven accuracy claims: The author explicitly states that results are not a real-world benchmark.
• Limited scope of use case: The tool is described as useful for postmortems but does not address broader software development workflows.
• Single-person team: Only one member (Dried Sandwich) is listed, suggesting limited capacity for scaling or iteration.
• No commercialization path: No pricing, monetization, or go-to-market strategy is evident.

Evidence The author’s own write-up.

Inference This appears to be a proof-of-concept or hackathon prototype, not a product ready for enterprise adoption.

1. What specific incident management workflows does RepoWitness aim to improve?
2. How would you scale this beyond the current synthetic testing environment?
3. Are there any plans to integrate with existing issue trackers or CI systems?
4. Has anyone outside of the author tested or used the tool in practice?
5. What is your roadmap for moving from prototype to production-ready software?