artifact

What problem does it solve? When an agent produces a valuable result (a log, an analysis, a query output), that result normally lives only in the conversation and must be regenerated by any follow-on agent. This Skill performs the requested work once and writes the full result to a distinct, self-contained artifact file so later agents can read it instead of redoing the work. ## Core Features & Use Cases - Per-request artifact files: Each invocation derives a kebab-case slug from the request (e.g. "log", "analyze", "summarize-failing-tests") and writes a separate timestamped file, so multiple calls in one turn never overwrite each other. - Predictable latest pointers: A per-slug artifact-{slug}-latest.md symlink gives follow-on agents a stable read path without knowing timestamps. - Chained artifacts: New artifacts can build on prior ones via a based_on frontmatter reference, e.g. an analysis artifact referencing an earlier log artifact. - Use Case: Run /artifact log the test output to capture a failing test run, then later /artifact analyze the output — the analysis reads the saved log artifact rather than rerunning the tests. ## Quick Start Ask the agent to run "artifact summarize the failing tests" and it will perform the summarization and save the result to a reusable artifact file.