cache-expert

The cache-expert skill provides deep technical insight into the Dagger Engine's caching architecture. Dagger treats build and deployment workflows as Directed Acyclic Graphs (DAGs) of operations on immutable values, where each operation is deterministic. By leveraging this immutability, the engine derives cache keys directly from inputs and operations. This skill is designed for engineers and platform developers who need to diagnose performance regressions, unexpected cache misses, or invalidation anomalies within complex pipeline structures. It serves as an interface for understanding how Dagger serializes DAGs into content-addressed IDs and how to interpret internal engine diagnostics.

• Expert-level analysis of the DAG model, immutable data structures, and deterministic operation execution.

• Diagnostic workflows for debugging cache invalidation logic and resolving unexpected cache misses in CI/CD environments.

• Guidance on interpreting the Dagger Engine internal cache state, including the use of tools like the dagql-cache-analyzer for analyzing engine snapshot dumps.

• Support for implementing or extending caching-related engine features based on the GraphQL API and internal engine state.

• Insights into how Dagger automates software delivery while ensuring repeatable builds via content-addressed caching.

• Intended for developers working with Dagger CLI and Engine internals, specifically those debugging CI pipelines or optimizing build performance.

• Use the provided dagql-cache-analyzer script to process /debug/dagql/cache snapshot dumps for offline inspection of retained roots and result categories.

• Primary inputs include engine debug logs, cache snapshot dumps, and pipeline operation traces; outputs include diagnostic summaries and performance optimization strategies.

• Note that documentation for internal engine mechanics may be experimental or out-of-date; the definitive source of truth is the Dagger source code and the references/debugging.md guide within the repository.

• Be aware that caching behavior is fundamentally tied to the input immutability; external side effects or non-deterministic inputs will lead to cache breaks.