What it does
Dev Artifact Parser is the adapter between the files your tools already produce and the agents that want structured input. You paste a raw artifact — an `lcov.info` from `vitest --coverage`, the output of `npm audit --json`, or a Terraform plan from `terraform show -json` — and it returns the exact records the matching agent consumes: coverage rows for test-coverage-gap, advisory rows for dependency-and-cve-upgrade, resource changes for iac-review. What makes it worth having as its own agent is that it refuses rather than guesses. Text it does not recognise comes back as `needs_input` with the supported formats listed, never a speculative parse. If you name a format that contradicts the content, the parse is refused instead of forced. And a record it cannot use is reported by name in the warnings — never dropped silently, and never replaced with sample data. `dataSource` is always `provided`, because there is no fixture path in the code at all. `explain_gaps` is the other half and the more useful one. After a parse it tells you what the target agent still needs that this artifact cannot supply — an LCOV file has no notion of business criticality, so it says so rather than letting you discover it downstream. That turns a partial input into a known-partial input. `detect_format` answers what a file is, and returns the markers that decided it. `list_supported` maps every format to the command that produces it and the agent that consumes it. No connectors and no credentials. It never reads a repository, calls a network API or touches the filesystem — everything it reports comes from the text in the call. Built for developers and CI owners who have the artifact and want the analysis, without hand-shaping JSON in between.
Example prompts
- Here is my coverage/lcov.info — turn it into input for the coverage agent
- Parse this npm audit --json output
- What artifact formats do you support?
